def test_outlier_influence_funcs():
#smoke test
x = add_constant(np.random.randn(10, 2))
y = x.sum(1) + np.random.randn(10)
res = OLS(y, x).fit()
oi.summary_table(res, alpha=0.05)
res2 = OLS(y, x[:,0]).fit()
oi.summary_table(res2, alpha=0.05)
infl = res2.get_influence()
infl.summary_table()
def test_outlier_influence_funcs():
# smoke test
x = add_constant(np.random.randn(10, 2))
y = x.sum(1) + np.random.randn(10)
res = OLS(y, x).fit()
oi.summary_table(res, alpha=0.05)
res2 = OLS(y, x[:, 0]).fit()
oi.summary_table(res2, alpha=0.05)
infl = res2.get_influence()
infl.summary_table()
def seasonal_prediction(df,
df_result,
y_name,
time_name,
new_t,
period,
show=False,
option='cma'):
"""
df should be the deseasoned df if possible
"""
y_v = df[y_name]
if len(new_t) == 0:
_, data, _ = sso.summary_table(df_result, alpha=0.05)
trend_proj = df_result.predict(sm.add_constant(new_t))
df_result.predict(sm.add_constant(new_t))
tdf = df.copy()
tdf[f'Pre_{y_name}'] = data[:, 2] * tdf['SeaIdx']
return tdf
else:
new_t = np.array(new_t)
SI, SIid = seasonal_index(y_v, period, show, option=option)
# des_df = smoothing_cma(df, y_name, time_name,
# period=period, show=show) # final df secured
trend_proj = df_result.predict(sm.add_constant(new_t))
seasonal_adj = trend_proj * SI
# new_t = np.arange(12, 16)
_, data, _ = sso.summary_table(df_result, alpha=0.05)
trend_proj = df_result.predict(sm.add_constant(new_t))
df_result.predict(sm.add_constant(new_t))
tdf = df.copy()
tdf[f'Pre_{y_name}'] = data[:, 2] * tdf['SeaIdx']
# tdf[x_name] = np.append(tdf[x_name], new_t)
for i, t in enumerate(new_t):
tdf = tdf.append(
{
time_name:
t,
'SID':
tdf['SID'].values[-(1 + i) - (len(new_t) - (1 + i))],
'SeaIdx':
tdf['SeaIdx'].values[-(1 + i) - (len(new_t) - (1 + i))],
f'Pre_{y_name}':
trend_proj[i] * tdf['SeaIdx'].values[-(1 + i) -
(len(new_t) -
(1 + i))]
},
ignore_index=True)
return tdf
def test_outlier_influence_funcs(reset_randomstate):
x = add_constant(np.random.randn(10, 2))
y = x.sum(1) + np.random.randn(10)
res = OLS(y, x).fit()
out_05 = oi.summary_table(res)
# GH3344 : Check alpha has an effect
out_01 = oi.summary_table(res, alpha=0.01)
assert_(np.all(out_01[1][:, 6] <= out_05[1][:, 6]))
assert_(np.all(out_01[1][:, 7] >= out_05[1][:, 7]))
res2 = OLS(y, x[:,0]).fit()
oi.summary_table(res2, alpha=0.05)
infl = res2.get_influence()
infl.summary_table()
def test_outlier_influence_funcs(reset_randomstate):
x = add_constant(np.random.randn(10, 2))
y = x.sum(1) + np.random.randn(10)
res = OLS(y, x).fit()
out_05 = oi.summary_table(res)
# GH3344 : Check alpha has an effect
out_01 = oi.summary_table(res, alpha=0.01)
assert_(np.all(out_01[1][:, 6] <= out_05[1][:, 6]))
assert_(np.all(out_01[1][:, 7] >= out_05[1][:, 7]))
res2 = OLS(y, x[:, 0]).fit()
oi.summary_table(res2, alpha=0.05)
infl = res2.get_influence()
infl.summary_table()
def OLS_fit(S1="600015", S2="600016"):
import statsmodels.api as sm
from statsmodels.stats.outliers_influence import summary_table
db = dbloader("./dataset/training_data")
ts1 = db.load_day_a(S1, "20090101", "20091130")["Closing Price"]
ts2 = db.load_day_a(S2, "20090101", "20091130")["Closing Price"]
ts1, ts2 = align_series(ts1, ts2)
x = ts1.values
Y = ts2.values
X = sm.add_constant(x)
res = sm.OLS(Y, X).fit()
print(res.summary())
_, data, _ = summary_table(res)
plt.plot(Y, label="real")
plt.plot(res.fittedvalues, label="fitted")
plt.legend()
plt.savefig("./image/OLS_{}_{}.png".format(S1, S2))
plt.show()
w1 = res.params[1]
diff = ts2 - w1 * ts1
diff_mean = diff.mean()
diff_std = diff.std()
mean_line = pd.Series(diff_mean, index=diff.index)
up_line = pd.Series(diff_mean + diff_std, index=diff.index)
down_line = pd.Series(diff_mean - diff_std, index=diff.index)
sets = pd.concat([diff, mean_line, up_line, down_line], axis=1)
sets.columns = ["diff", "mean", "up", "down"]
sets.plot(figsize=(14, 7))
plt.savefig("./image/OLS_diff_{}_{}.png".format(S1, S2), dpi=800)
plt.show()
def simple_CIPIINT_regplot(df, xname, yname, alpha=0.05):
print(
"\n\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n|CI PI Interval plot - simple|\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n"
)
print("using alpha = ", alpha)
df_sorted = df.sort_values([xname])
result = smf.ols(yname + '~' + xname, data=df_sorted).fit()
x = df_sorted[xname].values
y = df_sorted[yname].values
st, data, ss2 = sso.summary_table(result, alpha=alpha)
fittedvalues = data[:, 2]
predict_mean_se = data[:, 3]
predict_mean_ci_low, predict_mean_ci_upp = data[:, 4:6].T
predict_ci_low, predict_ci_upp = data[:, 6:8].T
plt.plot(x, y, 'o', color='gray')
plt.plot(x, fittedvalues, '-', lw=0.5)
plt.plot(x, predict_mean_ci_low, 'r-', lw=0.4)
plt.plot(x, predict_mean_ci_upp, 'r-', lw=0.4)
plt.plot(x, predict_ci_low, 'b--', lw=0.4)
plt.plot(x, predict_ci_upp, 'b--', lw=0.4)
plt.title('CI PI plot')
plt.xlabel(xname)
plt.ylabel(yname)
plt.legend([
'data points', 'regression model', 'confidence interval',
'prediction interval'
],
title='Legends',
bbox_to_anchor=(1.3, 1),
prop={'size': 6})
plt.show()
def multiple_durbin_watson(df, xnames, yname, alpha=0.05):
print("\n\n========== Durbin-Watson ==========\n")
y_data = df[yname]
x_data_ar = []
for i in range(len(xnames)):
x_data_ar.append(df[xnames[i]])
x_data_ar = np.asarray(x_data_ar)
x_data_T = x_data_ar.T
x_data = pd.DataFrame(x_data_T, columns=xnames)
x_data2 = sm.add_constant(x_data)
olsmod = sm.OLS(y_data, x_data2)
result = olsmod.fit()
st, data, ss2 = sso.summary_table(result, alpha=alpha)
print("Columns in data are: %s" % ss2)
# Predicted value
y_pre = data[:, 2]
# Studentized Residual
SD = data[:, 10]
x_square_sum = np.vdot(SD, SD)
print("x_square_sum = ", x_square_sum)
size = SD.size
print("size = ", size)
x_d = np.zeros((size))
print("x_d = ", x_d)
l_size = size - 1
for i in range(l_size):
x_d[i + 1] = SD[i + 1] - SD[i]
print("x_d = ", x_d)
d = np.vdot(x_d, x_d) / x_square_sum
print("d = ", d)
def regress(x,y,alpha=.05,xlabel='',ylabel='',title=''):
if x.name=='':
x.name='x'
X = sm.add_constant(x)
res = sm.OLS(y, X).fit()
st, data, ss2 = summary_table(res, alpha=0.05)
fittedvalues = data[:,2]
#predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
fig, ax = plt.subplots(figsize=(8,6))
ax.scatter(x, y, label="data",s=5)
ax.plot(x, fittedvalues, 'r-', label='OLS: 95% confidence')
ax.plot(x, predict_ci_low, 'b--')
ax.plot(x, predict_ci_upp, 'b--')
ax.plot(x, predict_mean_ci_low, 'g--')
ax.plot(x, predict_mean_ci_upp, 'g--')
ax.scatter(x=x.tail(1), y=y.tail(1), color='Red', s=25, label="Now")
ax.legend(loc='best');
equation = ylabel+" = %.4f"%res.params[1] +" * " + xlabel +" + " + "%.4f"%res.params[0]
ax.set_xlabel(xlabel + " "+ equation, color='g',fontsize = 15);
ax.set_ylabel(ylabel, color='b',fontsize = 15);
plt.show()
fig, ax = plt.subplots(figsize=(8,6))
plt.plot(x.index,data[:,8])
plt.title("Residual plot: "+equation,fontsize=20)
ax.set_xlabel("Date ", color='g',fontsize = 15);
ax.set_ylabel("Residual ", color='g',fontsize = 15);
plt.show()
return data
def linear_regression(self, x, y):
x = sm.add_constant(x)
regr = sm.OLS(y, x)
res = regr.fit()
# Get fitted values from model to plot
st, data, ss2 = summary_table(res, alpha=0.05)
fitted_values = data[:, 2]
return fitted_values
def figplot(x, y, sims, clrs, xlab, ylab, fig, n):
fig.add_subplot(2, 2, n)
y2 = list(y)
x2 = list(x)
clrs = list(clrs)
plt.scatter(x2, y2, color=clrs, s=2, linewidths=0.0)
d = pd.DataFrame({'x': list(x2)})
d['y'] = list(y2)
f = smf.ols('y ~ x', d).fit()
st, data, ss2 = summary_table(f, alpha=0.05)
fitted = data[:, 2]
m, b, r, p, std_err = stats.linregress(x2, y2)
if n == 1:
lab = r'$R_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$R_{microbes}$' + ' = 2.34*' + r'$N$' + '$^{0.14}$' + '\n'
lab += r'$R_{macrobes}$' + ' = 1.7*' + r'$N$' + '$^{0.11}$'
plt.text(0.2, 1.4, lab, fontsize=7)
elif n == 2:
lab = r'$D_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$D_{microbes}$' + ' = 0.44*' + r'$N$' + '$^{0.92}$' + '\n'
lab += r'$D_{macrobes}$' + ' = 0.23*' + r'$N$' + '$^{0.99}$'
plt.text(0.2, 2.5, lab, fontsize=7)
elif n == 3:
lab = r'$E_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$E_{microbes}$' + ' = 0.58*' + r'$N$' + '$^{-0.23}$' + '\n'
lab += r'$E_{macrobes}$' + ' = 1.15*' + r'$N$' + '$^{-0.21}$'
plt.text(0.2, -3.4, lab, fontsize=7)
elif n == 4:
lab = r'$S_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$S_{microbes}$' + ' = 1.77*' + r'$N$' + '$^{0.38}$' + '\n'
lab += r'$S_{macrobes}$' + ' = 1.77*' + r'$N$' + '$^{0.24}$'
plt.text(0.2, 2.8, lab, fontsize=7)
if n == 3: plt.legend(loc='best', fontsize=6, frameon=False)
plt.plot(x2, fitted, color='k', ls='--', lw=1.0, alpha=0.9)
plt.xlabel(xlab, fontsize=8)
plt.ylabel(ylab, fontsize=8)
plt.tick_params(axis='both', labelsize=5)
plt.xlim(0, 1.05 * max(x2))
if n == 1: plt.ylim(0.0, max(y2))
elif n == 2: plt.ylim(0.0, max(y2))
elif n == 3: plt.ylim(min(y2), 0)
elif n == 4: plt.ylim(0.4, max(y2))
return fig
def linearR(df, column1, column2):
x = sm.add_constant(df.toPandas()[column1])
y = df.toPandas()[column2]
regr = sm.OLS(y, x)
res = regr.fit()
st, data, ss2 = summary_table(res, alpha=0.05)
fitted_values = data[:, 2]
return fitted_values
def figplot(clrs, x, y, xlab, ylab, fig, n):
fig.add_subplot(2, 2, n)
plt.xscale('log')
if n == 1: plt.yscale('log', subsy=[1, 2])
plt.yscale('log')
plt.minorticks_off()
d = pd.DataFrame({'x': np.log10(x)})
d['y'] = np.log10(y)
f = smf.ols('y ~ x', d).fit()
m, b, r, p, std_err = stats.linregress(np.log10(x), np.log10(y))
st, data, ss2 = summary_table(f, alpha=0.05)
fitted = data[:, 2]
mean_ci_low, mean_ci_upp = data[:, 4:6].T
ci_low, ci_upp = data[:, 6:8].T
x, y, fitted, ci_low, ci_upp, clrs = zip(
*sorted(zip(x, y, fitted, ci_low, ci_upp, clrs)))
x = np.array(x)
y = np.array(y)
fitted = 10**np.array(fitted)
ci_low = 10**np.array(ci_low)
ci_upp = 10**np.array(ci_upp)
if n == 1:
lbl = r'$rarity$' + ' = ' + str(round(
10**b, 1)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$'
elif n == 2:
lbl = r'$Nmax$' + ' = ' + str(round(
10**b, 1)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$'
elif n == 3:
lbl = r'$Ev$' + ' = ' + str(round(
10**b, 1)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$'
elif n == 4:
lbl = r'$S$' + ' = ' + str(round(
10**b, 1)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$'
plt.scatter(x, y, s=sz, color=clrs, linewidths=0.0, edgecolor=None)
plt.fill_between(x, ci_upp, ci_low, color='0.5', lw=0.1, alpha=0.2)
plt.plot(x, fitted, color='k', ls='--', lw=0.5, label=lbl)
if n == 3: plt.legend(loc=3, fontsize=8, frameon=False)
else: plt.legend(loc=2, fontsize=8, frameon=False)
plt.xlabel(xlab, fontsize=10)
plt.ylabel(ylab, fontsize=10)
plt.tick_params(axis='both', labelsize=6)
if n in [2, 4]: plt.ylim(min(y), max(y))
elif n == 1: plt.ylim(min(ci_low), max(ci_upp))
elif n == 3: plt.ylim(0.1, 1.1)
return fig
def fixed_effect(treat, control, k, SEED):
x = np.arange(T)
x = np.concatenate([x for _ in range(N_tr+N_co)]).reshape(-1,1)
units = np.concatenate([[i for _ in range(T)] for i in range(N_tr+N_co)]).reshape(-1,1)
treated = np.logical_and((units<N_tr), (x>=T0)).astype("float")
y = np.concatenate([treat.reshape(-1,1),control.reshape(-1,1)])
COLUMNS = ["time", "y", "unit", "treated"]
data = pd.DataFrame(np.concatenate((x,y,units,treated),axis=1),columns=COLUMNS)
data.to_csv(synthetic_path+"/data_{}.csv".format(SEED), index=False)
return
fit = ols('y ~ 1 + C(time) + C(unit) + treated:C(time)', data=data).fit()
ypred = fit.predict(data)
m_tr = ypred[:N_tr*T].to_numpy().reshape(N_tr,T)
m_co = ypred[N_tr*T:].to_numpy().reshape(N_co,T)
# print(fit.summary())
for t in range(T0, T, 1):
m_tr[:, t] -= fit.params["treated:C(time)[{}.0]".format(t)]
_, data, _ = summary_table(fit, alpha=0.05)
predict_mean_ci_lower, predict_mean_ci_upper = data[:, 4:6].T
lower_tr = predict_mean_ci_lower[:N_tr*T].reshape(N_tr,T)
upper_tr = predict_mean_ci_upper[:N_tr*T].reshape(N_tr,T)
lower_co = predict_mean_ci_lower[N_tr*T:].reshape(N_co,T)
upper_co = predict_mean_ci_upper[N_tr*T:].reshape(N_co,T)
for t in range(T0, T, 1):
lower_tr[:, t] -= fit.conf_int().loc["treated:C(time)[{}.0]".format(t),1]
upper_tr[:, t] -= fit.conf_int().loc["treated:C(time)[{}.0]".format(t),0]
test_t = np.arange(T)
# plt.plot(test_t, np.mean(control, axis=0), color='grey', alpha=0.8)
# plt.plot(test_t, np.mean(m_co, axis=0), 'k--', linewidth=1.0, label='Estimated Y(0)')
# plt.fill_between(test_t, np.mean(lower_co, axis=0), np.mean(upper_co, axis=0), alpha=0.5)
# plt.show()
ATT = np.stack([np.mean(treat-m_tr, axis=0),
np.mean(treat-upper_tr, axis=0),
np.mean(treat-lower_tr, axis=0)])
plt.rcParams["figure.figsize"] = (15,5)
plt.plot(test_t, ATT[0],'k--', linewidth=1.0, label="Estimated ATT")
plt.fill_between(test_t, ATT[1], ATT[2], alpha=0.5, label="ATT 95% CI")
plt.legend(loc=2)
plt.savefig(synthetic_path+"/fixedeffect{}_{}.png".format(k, SEED))
plt.close()
np.savetxt(synthetic_path+"/fixedeffect{}_{}.csv".format(k, SEED), ATT, delimiter=",")
def figplot(x, y, xlab, ylab, fig, n, binned=1):
'''main figure plotting function'''
fig.add_subplot(3, 3, n)
x = np.log10(x)
y = np.log10(y)
y2 = list(y)
x2 = list(x)
if binned == 1:
X, Y = (np.array(t) for t in zip(*sorted(zip(x2, y2))))
Xi = xfrm(X, max(X) * 1.05)
bins = np.linspace(np.min(Xi), np.max(Xi) + 1, 100)
ii = np.digitize(Xi, bins)
y2 = np.array([
np.mean(Y[ii == i]) for i in range(1, len(bins))
if len(Y[ii == i]) > 0
])
x2 = np.array([
np.mean(X[ii == i]) for i in range(1, len(bins))
if len(X[ii == i]) > 0
])
d = pd.DataFrame({'size': list(x2)})
d['rate'] = list(y2)
f = smf.ols('rate ~ size', d).fit()
coef = f.params[1]
st, data, ss2 = summary_table(f, alpha=0.05)
fitted = data[:, 2]
mean_ci_low, mean_ci_upp = data[:, 4:6].T
ci_low, ci_upp = data[:, 6:8].T
x2, y2, fitted, ci_low, ci_upp = zip(
*sorted(zip(x2, y2, fitted, ci_low, ci_upp)))
plt.scatter(x2,
y2,
color='SkyBlue',
alpha=1,
s=12,
linewidths=0.5,
edgecolor='Steelblue')
plt.fill_between(x2, ci_upp, ci_low, color='b', lw=0.1, alpha=0.15)
plt.plot(x2, fitted, color='b', ls='--', lw=1.0, alpha=0.9)
plt.xlabel(xlab, fontsize=10)
plt.ylabel(ylab, fontsize=10)
plt.tick_params(axis='both', labelsize=6)
plt.xlim(0.9 * min(x2), 1.1 * max(x2))
plt.ylim(min(ci_low), max(ci_upp))
plt.title('$z$ = ' + str(round(coef, 2)), fontsize=10)
return fig
def ciAnalysis(re,x,y): st, data, ss2 = summary_table(re, alpha=0.10) fittedvalues = data[:,2] predict_mean_se = data[:,3] predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T predict_ci_low, predict_ci_upp = data[:,6:8].T plt.plot(x, y, 'o') plt.plot(x, fittedvalues, '-', lw=2) plt.plot(x, predict_ci_low, 'r--', lw=2) plt.plot(x, predict_ci_upp, 'r--', lw=2) plt.plot(x, predict_mean_ci_low, 'r--', lw=2) plt.plot(x, predict_mean_ci_upp, 'r--', lw=2) plt.show()
def seasonal_index(y_v, period: int, show=False, option='cma'):
"""
return seasonal index and a Series with all observations linked to its seasonal index
"""
n = period
if option == 'cma':
SI_MA_a = np.zeros(len(y_v))
SI_MA_a[:] = np.nan
SI_MA_a = y_v / cma(y_v, period)
SI_id_s = np.arange(1, len(y_v) + 1)
SI_id = SI_id_s - np.floor(SI_id_s / n) * n
SI_id[np.where((SI_id[:] == 0))] = n
SI_MA_df = pd.DataFrame({'SIMA': SI_MA_a, 'SIid': SI_id})
SI_MA_u = np.zeros(n)
for j in range(1, n + 1):
SI_MA_u[j - 1] = SI_MA_df['SIMA'][SI_MA_df['SIid'] ==
j].dropna().mean()
SI_MA = SI_MA_u / sum(SI_MA_u) * n
if show:
print('Seasonal Index:', SI_MA)
return SI_MA, SI_MA_df['SIid']
elif option == 'lr':
y_data = y_v
X_data_ar = np.arange(1, len(y_v) + 1)
X_data_T = X_data_ar.T
X_data = pd.DataFrame(X_data_T, columns=['Time'])
X_data = sm.add_constant(X_data)
olsmod = sm.OLS(y_data, X_data)
result_reg = olsmod.fit()
st, data, ss2 = sso.summary_table(result_reg, alpha=0.05)
y_v_LR_a = data[:, 2]
SI_LR_a = y_v / y_v_LR_a
SI_id_s = np.arange(1, len(y_v) + 1)
SI_id = SI_id_s - np.floor(SI_id_s / n) * n
SI_id[np.where((SI_id[:] == 0))] = n
SI_LR_a_df = pd.DataFrame({'SILR': SI_LR_a, 'SIid': SI_id})
SI_LR_u = np.zeros(n)
for j in range(1, n + 1):
SI_LR_u[j - 1] = SI_LR_a_df['SILR'][SI_LR_a_df['SIid'] ==
j].dropna().mean()
SI_LR = SI_LR_u / sum(SI_LR_u) * n
if show:
print('Seasonal Index:', SI_LR)
return SI_LR, SI_LR_a_df['SIid']
def figplot(clrs, x, y, xlab, ylab, fig, n):
fig.add_subplot(2, 2, n)
plt.xscale('log')
if n == 1: plt.yscale('log', subsy=[1, 2])
plt.yscale('log')
plt.minorticks_off()
d = pd.DataFrame({'x': np.log10(x)})
d['y'] = np.log10(y)
f = smf.ols('y ~ x', d).fit()
m, b, r, p, std_err = stats.linregress(np.log10(x), np.log10(y))
st, data, ss2 = summary_table(f, alpha=0.05)
fitted = data[:,2]
mean_ci_low, mean_ci_upp = data[:,4:6].T
ci_low, ci_upp = data[:,6:8].T
x, y, fitted, ci_low, ci_upp, clrs = zip(*sorted(zip(x, y, fitted, ci_low, ci_upp, clrs)))
x = np.array(x)
y = np.array(y)
fitted = 10**np.array(fitted)
ci_low = 10**np.array(ci_low)
ci_upp = 10**np.array(ci_upp)
if n == 1: lbl = r'$rarity$'+ ' = '+str(round(10**b,1))+'*'+r'$N$'+'$^{'+str(round(m,2))+'}$'
elif n == 2: lbl = r'$Nmax$'+ ' = '+str(round(10**b,1))+'*'+r'$N$'+'$^{'+str(round(m,2))+'}$'
elif n == 3: lbl = r'$Ev$'+ ' = '+str(round(10**b,1))+'*'+r'$N$'+'$^{'+str(round(m,2))+'}$'
elif n == 4: lbl = r'$S$'+ ' = '+str(round(10**b,1))+'*'+r'$N$'+'$^{'+str(round(m,2))+'}$'
plt.scatter(x, y, s = sz, color=clrs, linewidths=0.0, edgecolor=None)
plt.fill_between(x, ci_upp, ci_low, color='0.5', lw=0.1, alpha=0.2)
plt.plot(x, fitted, color='k', ls='--', lw=0.5, label = lbl)
if n == 3: plt.legend(loc=3, fontsize=8, frameon=False)
else: plt.legend(loc=2, fontsize=8, frameon=False)
plt.xlabel(xlab, fontsize=10)
plt.ylabel(ylab, fontsize=10)
plt.tick_params(axis='both', labelsize=6)
if n in [2, 4]: plt.ylim(min(y), max(y))
elif n == 1: plt.ylim(min(ci_low), max(ci_upp))
elif n == 3: plt.ylim(0.1, 1.1)
return fig
def lm(x, y, alpha=ALPHA):
"fits an OLS from statsmodels. returns tuple."
x, y = map(plot_friendly, [x, y])
if _isdate(x[0]):
x = np.array([i.toordinal() for i in x])
X = sm.add_constant(x)
fit = sm.OLS(y, X).fit()
prstd, iv_l, iv_u = wls_prediction_std(fit)
_, summary_values, summary_names = summary_table(fit, alpha=alpha)
df = pd.DataFrame(summary_values, columns=map(snakify, summary_names))
fittedvalues = df['predicted_value']
predict_mean_se = df['std_error_mean_predict']
predict_mean_ci_low = df['mean_ci_95%_low']
predict_mean_ci_upp = df['mean_ci_95%_upp']
predict_ci_low = df['predict_ci_95%_low']
predict_ci_upp = df['predict_ci_95%_upp']
return (fittedvalues, predict_mean_ci_low, predict_mean_ci_upp)
def lm(x, y, alpha=ALPHA):
"fits an OLS from statsmodels. returns tuple."
x, y = map(plot_friendly, [x,y])
if _isdate(x[0]):
x = np.array([i.toordinal() for i in x])
X = sm.add_constant(x)
fit = sm.OLS(y, X).fit()
prstd, iv_l, iv_u = wls_prediction_std(fit)
_, summary_values, summary_names = summary_table(fit, alpha=alpha)
df = pd.DataFrame(summary_values, columns=map(snakify, summary_names))
fittedvalues = df['predicted_value']
predict_mean_se = df['std_error_mean_predict']
predict_mean_ci_low = df['mean_ci_95%_low']
predict_mean_ci_upp = df['mean_ci_95%_upp']
predict_ci_low = df['predict_ci_95%_low']
predict_ci_upp = df['predict_ci_95%_upp']
return (fittedvalues, predict_mean_ci_low, predict_mean_ci_upp)
def _fit_reg(fit_reg, ci, ax, x, y, data, color, line_kws):
if not fit_reg:
return None
if ci is None:
ci = 0
if ci < 0 or ci >= 100:
raise ValueError('ci must be between 0 and 100 or `None`')
if line_kws is None:
line_kws = {}
if 'lw' not in line_kws:
line_kws['lw'] = 3
X = data[x].values
if len(X) == 1:
return None
idx_order = X.argsort()
y = data[y].values
if len(X) == 2:
ax.plot(X, y, color=color, **line_kws)
return None
X = sm.add_constant(X)
# if all x's are the same value, there can be no regression line
if X.shape[1] == 1:
return 1
ols = sm.OLS(y, X).fit()
pred_obj = ols.get_prediction()
pred = pred_obj.predicted_mean[idx_order]
try:
ax.plot(X[idx_order, 1], pred, color=color, **line_kws)
except IndexError:
print(f"col is {x}")
print(X.shape)
print(data[x].values)
print(X)
if ci != 0:
st, data, ss2 = summary_table(ols, alpha=1 - ci / 100)
ax.fill_between(X[idx_order, 1],
data[idx_order, 4],
data[idx_order, 5],
alpha=.3,
color=color)
def simple_outliers_DIY(df, xname, yname, alpha=0.05):
# Fit regression model
result = smf.ols(yname + '~' + xname, data=df).fit()
# studentized residual
st1, data1, ss3 = sso.summary_table(result, alpha=alpha)
Residual = data1[:, 8]
STD_Residual = data1[:, 10]
mu = np.mean(STD_Residual)
sigma = np.std(STD_Residual)
print("◆ Outliers Finding\n")
print("(remove by yourself!)\n")
df_out = pd.DataFrame(STD_Residual, columns=['SD'])
filter = (df_out['SD'] < -2) | (df_out['SD'] > 2)
print("Outliers by SD = ")
print(df_out['SD'].loc[filter])
print("\nActual ID: ", df_out['SD'].loc[filter].index + 1)
return df_out['SD'].loc[filter].index
def lm(x, y, alpha=ALPHA):
"fits an OLS from statsmodels. returns tuple."
x_is_date = _isdate(x.iloc[0])
if x_is_date:
x = np.array([i.toordinal() for i in x])
X = sm.add_constant(x)
fit = sm.OLS(y, X).fit()
prstd, iv_l, iv_u = wls_prediction_std(fit)
_, summary_values, summary_names = summary_table(fit, alpha=alpha)
df = pd.DataFrame(summary_values, columns=map(_snakify, summary_names))
# TODO: indexing w/ data frame is messing everything up
fittedvalues = df['predicted_value'].values
predict_mean_ci_low = df['mean_ci_95%_low'].values
predict_mean_ci_upp = df['mean_ci_95%_upp'].values
predict_ci_low = df['predict_ci_95%_low'].values
predict_ci_upp = df['predict_ci_95%_upp'].values
if x_is_date:
x = [Timestamp.fromordinal(int(i)) for i in x]
return (x, fittedvalues, predict_mean_ci_low, predict_mean_ci_upp)
def lm(x, y, alpha=ALPHA):
"fits an OLS from statsmodels. returns tuple."
x_is_date = _isdate(x.iloc[0])
if x_is_date:
x = np.array([i.toordinal() for i in x])
X = sm.add_constant(x)
fit = sm.OLS(y, X).fit()
prstd, iv_l, iv_u = wls_prediction_std(fit)
_, summary_values, summary_names = summary_table(fit, alpha=alpha)
df = pd.DataFrame(summary_values, columns=map(_snakify, summary_names))
# TODO: indexing w/ data frame is messing everything up
fittedvalues = df['predicted_value'].values
predict_mean_ci_low = df['mean_ci_95%_low'].values
predict_mean_ci_upp = df['mean_ci_95%_upp'].values
predict_ci_low = df['predict_ci_95%_low'].values
predict_ci_upp = df['predict_ci_95%_upp'].values
if x_is_date:
x = [Timestamp.fromordinal(int(i)) for i in x]
return (x, fittedvalues, predict_mean_ci_low, predict_mean_ci_upp)
def lm(x, y, alpha=ALPHA):
"fits an OLS from statsmodels. returns tuple."
import statsmodels.api as sm
from statsmodels.sandbox.regression.predstd import wls_prediction_std
from statsmodels.stats.outliers_influence import summary_table
x, y = map(plot_friendly, [x,y])
if _isdate(x[0]):
x = np.array([i.toordinal() for i in x])
X = sm.add_constant(x)
fit = sm.OLS(y, X).fit()
prstd, iv_l, iv_u = wls_prediction_std(fit)
_, summary_values, summary_names = summary_table(fit, alpha=alpha)
df = pd.DataFrame(summary_values, columns=map(snakify, summary_names))
fittedvalues = df['predicted_value']
predict_mean_se = df['std_error_mean_predict']
predict_mean_ci_low = df['mean_ci_95%_low']
predict_mean_ci_upp = df['mean_ci_95%_upp']
predict_ci_low = df['predict_ci_95%_low']
predict_ci_upp = df['predict_ci_95%_upp']
return (x, fittedvalues.tolist(), predict_mean_ci_low.tolist(),
predict_mean_ci_upp.tolist())
def get_pred_interval_sm(y, X, dfx, pi = 0.95):
import statsmodels.api as sm
from statsmodels.sandbox.regression.predstd import wls_prediction_std
from statsmodels.stats.outliers_influence import summary_table
df = dfx.copy()
Y = df[y]
X = df[X]
X = sm.add_constant(X)
re = sm.OLS(Y, X).fit()
print(re.summary())
st, data, ss2 = summary_table(re, alpha=1-pi)
fittedvalues = data[:, 2]
predict_mean_se = data[:, 3]
predict_mean_ci_low, predict_mean_ci_upp = data[:, 4:6].T
predict_ci_low, predict_ci_upp = data[:, 6:8].T
return predict_ci_low, fittedvalues, predict_ci_upp
def compare_two_tickers(TICKER_A, TICKER_B):
start = datetime.datetime(2017, 5, 1)
end = datetime.datetime(2018, 5, 27)
df = pd.DataFrame(columns=(TICKER_A, TICKER_B))
df[TICKER_A] = retrieve_ticker_prices(TICKER_A, start, end)
df[TICKER_B] = retrieve_ticker_prices(TICKER_B, start, end)
print(df)
# Plot the two time series
plot_price_series(df, TICKER_A, TICKER_B, start, end)
# Display a scatter plot of the two time series
plot_scatter_series(df, TICKER_A, TICKER_B)
# Calculate optimal hedge ratio "beta"
res = sm.OLS(endog=df[TICKER_B], exog=df[TICKER_A]).fit()
st, data, ss2 = summary_table(
res, alpha=0.05) # 置信水平alpha=5%，st数据汇总，data数据详情，ss2数据列名
# beta_hr = data[:2] # 等价于res.fittedvalues
beta_hr = res.params[TICKER_A]
# beta_hr = res.fittedvalues #获取拟合y值
# res.params # 拟合回归模型参数
# res.params[0] + res.params[1] * daily_data['temp'] == res.fittedvalues # 验证二维回归模型的拟合y值计算原理
# Calculate the residuals of the linear combination
df["res"] = df[TICKER_A] - beta_hr * df[TICKER_B]
print('===============')
print(beta_hr)
print('===============')
print(df['res'])
# Plot the residuals
plot_residuals(df, start, end)
# Calculate and output the CADF test on the residuals
calculate_adf(df['res'])
def drawLinearRegressionByTail(x, y, alpha, ax):
x = np.array(x)
res = sm.OLS(y, x).fit()
st, data, ss2 = summary_table(res, alpha=alpha)
fittedvalues = data[:, 2]
predict_mean_ci_low, predict_mean_ci_upp = data[:, 4:6].T
predict_ci_low, predict_ci_upp = data[:, 6:8].T
ax.plot(x, fittedvalues, 'r-', label='Параметрична пряма')
ax.plot(x,
predict_ci_low,
'black',
label='Толерантні межі',
linestyle='dashed')
ax.plot(x, predict_ci_upp, 'black', linestyle='dashed')
ax.plot(x,
predict_mean_ci_low,
'gray',
label='Довірчий інтервал',
linestyle='dashed')
ax.plot(x, predict_mean_ci_upp, 'gray', linestyle='dashed')
ax.legend(loc='best', fontsize='x-small')
def get_regression_summary(result,
conf_int=0.95,
columns=None,
as_dataframe=False):
column_mapper = {
'Obs': 'obs',
'Dep Var\nPopulation': 'dep_var_population',
'Predicted\nValue': 'predicted_value',
'Std Error\nMean Predict': 'std_error_mean_predict',
'Mean ci\n95% low': 'mean_ci_lo',
'Mean ci\n95% upp': 'mean_ci_up',
'Predict ci\n95% low': 'pred_ci_lo',
'Predict ci\n95% upp': 'pred_ci_up',
'Residual': 'residual',
'Std Error\nResidual': 'std_error_residual',
'Student\nResidual': 'student_residual',
"Cook's\nD": 'cooks'
}
# columns = columns or ['Mean ci\n95% low', 'Mean ci\n95% upp']
simple_table, data_table, table_columns = summary_table(result,
alpha=conf_int)
table_columns = [column_mapper.get(c) for c in table_columns]
if as_dataframe:
return pd.DataFrame(data_table, columns=table_columns)
def calcErrDistBug(in_filename,gold_in_filename,out_filename,title):
errs = []
abs_errs = []
mses = []
# read data
in_file = open(in_filename,'rb')
errs = cPickle.load(in_file)
abs_errs = cPickle.load(in_file)
mses = cPickle.load(in_file)
in_file.close()
g_in_file = open(gold_in_filename,'rb')
g_errs = cPickle.load(g_in_file)
g_abs_errs = cPickle.load(g_in_file)
g_mses = cPickle.load(g_in_file)
g_in_file.close()
# ensure that we don't actually take the log of 0
# FIXME: we may need to make this more dynamic so that it doesn't skew results.
g_mses = np.array(g_mses)
g_abs_errs = np.array(g_abs_errs)
zeros = g_mses>0.0
g_mses = g_mses[zeros]
g_abs_errs = g_abs_errs[zeros]
#zeros = g_abs_errs>0.0
zeros = g_abs_errs>1.0E-14
g_mses = g_mses[zeros]
g_abs_errs = g_abs_errs[zeros]
#print g_mses
#print g_abs_errs
#print g_mses
mses = np.array(mses)
abs_errs = np.array(abs_errs)
zeros = mses>0.0
mses = mses[zeros]
abs_errs = abs_errs[zeros]
#zeros = abs_errs>0.0
zeros = abs_errs>1.0E-14
mses = mses[zeros]
abs_errs = abs_errs[zeros]
skip = 5
mses = mses[::skip]
abs_errs = abs_errs[::skip]
#print g_mses
#print np.log2(g_mses)
g_dist = np.divide(g_mses,g_abs_errs)
dist = np.divide(mses,abs_errs)
# determine Ordinary Least Squares
X = np.log2(g_abs_errs)
X = sm.add_constant(X)
#print X
#print len(g_abs_errs)
#model = sm.OLS(g_mses,X)
model = sm.OLS(np.log2(g_mses),X)
#model = sm.RLM(np.log2(g_mses),X)
results = model.fit()
#print results.params
#print results.summary()
#print dir(results)
#print results.outlier_test()
prstd, iv_l, iv_u = wls_prediction_std(results)
st, data, ss2 = summary_table(results, alpha=0.05)
#print oi.OLSInfluence(results).influence
fittedvalues = data[:,2]
predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
# check we got the right things
#print np.max(np.abs(results.fittedvalues - fittedvalues))
#print np.max(np.abs(iv_l - predict_ci_low))
#print np.max(np.abs(iv_u - predict_ci_upp))
#legend = []
#legend.append('> to < Mutation')
#plt.loglog(abs_errs, mses, basex=2, linestyle='', marker='+', color='b')
#plt.plot(np.log2(abs_errs), np.log2(mses), linestyle='', marker='+', color='b')
#legend.append('> to < Mutation -- outliers')
o_x,o_y = returnOutliers(results,sm.add_constant(np.log2(abs_errs)),np.log2(mses),alpha=0.05)
#plt.plot(o_x, o_y, linestyle='', marker='*', color='b')
#plt.plot(abs_errs, dist, linestyle='', marker='o', color='b')
#print 'Mutation outliers: ' + `len(o_x)`
#legend.append('Bug-free')
#plt.loglog(g_abs_errs, g_mses, basex=2, linestyle='', marker='+', color='r')
#plt.plot(np.log2(g_abs_errs), np.log2(g_mses), linestyle='', marker='+', color='r')
#g_o_x,g_o_y = returnOutliers(results,sm.add_constant(np.log2(g_abs_errs)),np.log2(g_mses),alpha=0.05)
#plt.plot(g_o_x, g_o_y, linestyle='', marker='*', color='r')
#legend.append('Bug-free -- MSE/ABS')
#plt.plot(g_abs_errs, g_dist, linestyle='', marker='o', color='r')
#print 'Gold outliers: ' + `len(g_o_x)`
#legend.append('95% CI -')
#plt.loglog(g_abs_errs, iv_l, basex=2, linestyle='-', color='c')
#plt.plot(np.log2(g_abs_errs), iv_l, linestyle='-', color='c')
#legend.append('95% CI - manual')
#plt.plot(np.log2(g_abs_errs), predict_ci_low, linestyle='-', color='m')
#plt.loglog(g_abs_errs, results.fittedvalues, basex=2, linestyle='-', color='k')
#plt.plot(np.log2(g_abs_errs), results.fittedvalues, linestyle='-', color='k')
#legend.append('95% CI +')
#plt.loglog(g_abs_errs, iv_u, basex=2, linestyle='-', color='g')
#plt.plot(np.log2(g_abs_errs), iv_u, linestyle='-', color='g')
#legend.append('95% CI + manual')
#plt.plot(np.log2(g_abs_errs), predict_ci_upp, linestyle='-', color='y')
#leg = plt.legend(legend, 'lower right',ncol=1)
# fix up plotting to look nice
#plt.suptitle(title, fontsize=35)
#plt.xlabel('Sum of Absolute Execution Errors', fontsize=23)
#plt.ylabel('Mean Squared Error of Output', fontsize=23)
#plt.show()
return len(o_x),len(mses)
def Fig1(ref, Ones):
datasets = []
if ref == 'ClosedRef': GoodNames = ['EMPclosed', 'HMP', 'BIGN', 'TARA', 'BOVINE', 'HUMAN', 'LAUB', 'SED', 'CHU', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'] # all microbe data is MGRAST
if ref == 'OpenRef': GoodNames = ['EMPopen', 'HMP', 'BIGN', 'TARA', 'BOVINE', 'HUMAN', 'LAUB', 'SED', 'CHU', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'] # all microbe data is MGRAST
for name in os.listdir(mydir +'data/micro'):
if name in GoodNames: pass
else: continue
if Ones == 'N': path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y': path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'micro', num_lines])
print name, num_lines
for name in os.listdir(mydir +'data/macro'):
if name in GoodNames: pass
else: continue
if Ones == 'N': path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y': path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'macro', num_lines])
print name, num_lines
metrics = ['Nmax, '+r'$log_{10}$', 'McNaughton', 'Berger-Parker', 'Simpson\'s D']
fig = plt.figure()
for index, i in enumerate(metrics):
metric = i
fig.add_subplot(2, 2, index+1)
fs = 10 # font size used across figures
MicIntList, MicCoefList, MacIntList, MacCoefList, R2List, metlist = [[], [], [], [], [], []]
Nlist, Slist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList = [[], [], [], [], [], [], [], [], []]
EvarList, EQList, OList = [[],[],[]]
SimpDomList, McNList, LogSkewList, POnesList = [[],[],[],[]]
its = 10000
for n in range(its):
print n, metric
Nlist, Slist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList = [[], [], [], [], [], [], [], [], []]
EvarList, EQList, OList = [[],[],[]]
SimpDomList, McNList, LogSkewList, POnesList = [[],[],[],[]]
numMac = 0
numMic = 0
radDATA = []
for dataset in datasets:
name, kind, numlines = dataset
lines = []
small = ['BIGN', 'BOVINE', 'CHU', 'LAUB', 'SED']
big = ['HUMAN', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO']
if kind == 'macro':
lines = np.random.choice(range(1, numlines+1), 100, replace=True)
elif name in small:
lines = np.random.choice(range(1, numlines+1), 20, replace=True)
elif name in big:
lines = np.random.choice(range(1, numlines+1), 50, replace=True)
elif name == 'TARA':
lines = np.random.choice(range(1, numlines+1), 50, replace=True)
else:
lines = np.random.choice(range(1, numlines+1), 50, replace=True)
if Ones == 'N': path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y': path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
KindList.append(kind)
N = float(N)
S = float(S)
if S < 2 or N < 10: continue # Min species richness
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
# Dominance
BPlist.append(float(BP))
NmaxList.append(float(np.log10(float(Nmax))))
SimpDomList.append(float(SimpDom))
McNList.append(float(McN))
if kind == 'micro':
numMic += 1
klist.append('b')
if kind == 'macro':
klist.append('r')
numMac += 1
if index == 0: metlist = list(NmaxList)
elif index == 1: metlist = list(McNList)
elif index == 2: metlist = list(BPlist)
elif index == 3: metlist = list(SimpDomList)
# Multiple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
f = smf.ols('y ~ N * Kind', d).fit()
MacIntList.append(f.params[0])
MacCoefList.append(f.params[2])
if f.pvalues[1] < 0.05:
MicIntList.append(f.params[1] + f.params[0])
else:
MicIntList.append(f.params[0])
if f.pvalues[3] < 0.05:
MicCoefList.append(f.params[3] + f.params[2])
else:
MicCoefList.append(f.params[2])
R2List.append(f.rsquared)
MacListX = []
MacListY = []
MicListX = []
MicListY = []
for j, k in enumerate(KindList):
if k == 'micro':
MicListX.append(Nlist[j])
MicListY.append(metlist[j])
elif k == 'macro':
MacListX.append(Nlist[j])
MacListY.append(metlist[j])
MacPIx, MacFitted, MicPIx, MicFitted = [[],[],[],[]]
macCiH, macCiL, micCiH, micCiL = [[],[],[],[]]
lm = smf.ols('y ~ N * Kind', d).fit()
print metric, '\n', lm.summary()
f1 = smf.ols('y ~ N', d).fit()
print metric, '\n', f1.summary()
st, data, ss2 = summary_table(lm, alpha=0.05)
# ss2: Obs, Dep Var Population, Predicted Value, Std Error Mean Predict,
# Mean ci 95% low, Mean ci 95% upp, Predict ci 95% low, Predict ci 95% upp,
# Residual, Std Error Residual, Student Residual, Cook's D
fittedvalues = data[:,2]
predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
for j, kval in enumerate(KindList):
if kval == 'macro':
macCiH.append(predict_mean_ci_upp[j])
macCiL.append(predict_mean_ci_low[j])
MacPIx.append(Nlist[j])
MacFitted.append(f.fittedvalues[j])
elif kval == 'micro':
micCiH.append(predict_mean_ci_upp[j])
micCiL.append(predict_mean_ci_low[j])
MicPIx.append(Nlist[j])
MicFitted.append(f.fittedvalues[j])
MicPIx, MicFitted, micCiH, micCiL = zip(*sorted(zip(MicPIx, MicFitted, micCiH, micCiL)))
MacPIx, MacFitted, macCiH, macCiL = zip(*sorted(zip(MacPIx, MacFitted, macCiH, macCiL)))
for i in range(len(MicListX)):
plt.scatter(MacListX[i], MacListY[i], color = 'LightCoral', alpha= 1 , s = 4, linewidths=0.5, edgecolor='Crimson')
plt.scatter(MicListX[i], MicListY[i], color = 'SkyBlue', alpha= 1 , s = 4, linewidths=0.5, edgecolor='Steelblue')
plt.fill_between(MacPIx, macCiL, macCiH, color='r', lw=0.0, alpha=0.3)
plt.fill_between(MicPIx, micCiL, micCiH, color='b', lw=0.0, alpha=0.3)
MicInt = round(np.mean(MicIntList), 2)
MicCoef = round(np.mean(MicCoefList), 2)
MacInt = round(np.mean(MacIntList), 2)
MacCoef = round(np.mean(MacCoefList), 2)
r2 = round(np.mean(R2List), 2)
if index == 0:
plt.ylim(0, 6)
plt.xlim(1, 8)
plt.text(1.5, 5.3, r'$micro$'+ ' = '+str(round(MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(1.5, 4.7, r'$macro$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(1.5, 4.0, r'$R^2$' + '=' +str(round(r2,3)), fontsize=fs-1, color='k')
if index == 1:
plt.ylim(0, 120)
plt.xlim(1, 8)
#plt.text(4.0, 110, r'$micro$'+ ' = '+str(round(MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
#plt.text(4.0, 100, r'$macro$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(5.0, 90, r'$R^2$' + '=' +str(round(r2,3)), fontsize=fs-1, color='k')
if index == 2:
plt.ylim(0, 1.2)
plt.xlim(1, 8)
#plt.text(3.8, 1.10, r'$micro$'+ ' = '+str(round(MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
#plt.text(3.8, 1.0, r'$macro$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(1.5, 1.0, r'$R^2$' + '=' +str(round(r2,3)), fontsize=fs-1, color='k')
if index == 3:
plt.ylim(0, 1.3)
plt.xlim(1, 8)
#plt.text(1.5, 1.2, r'$micro$'+ ' = '+str(round(MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
#plt.text(1.5, 1.1, r'$macro$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(1.5, 1.1, r'$R^2$' + '=' +str(round(r2,3)), fontsize=fs-1, color='k')
plt.xlabel('Number of reads or individuals, '+ '$log$'+r'$_{10}$', fontsize=fs)
plt.ylabel(metric, fontsize=fs)
plt.tick_params(axis='both', which='major', labelsize=fs-1)
plt.subplots_adjust(wspace=0.4, hspace=0.4)
if ref == 'OpenRef'and Ones =='N': plt.savefig(mydir+'/figs/appendix/Dominance/SupplementaryDominanceFig-OpenRef_NoMicrobe1s.png', dpi=600, bbox_inches = "tight")
elif ref == 'OpenRef'and Ones =='Y': plt.savefig(mydir+'/figs/appendix/Dominance/SupplementaryDominanceFig-OpenRef.png', dpi=600, bbox_inches = "tight")
elif ref == 'ClosedRef'and Ones =='Y': plt.savefig(mydir+'/figs/appendix/Dominance/SupplementaryDominanceFig-ClosedRef.png', dpi=600, bbox_inches = "tight")
elif ref == 'ClosedRef'and Ones =='N': plt.savefig(mydir+'/figs/appendix/Dominance/SupplementaryDominanceFig-ClosedRef_NoMicrobe1s.png', dpi=600, bbox_inches = "tight")
#plt.show()
return
def Fig1(ref, Ones):
datasets = []
if ref == 'ClosedRef': GoodNames = ['MGRAST', 'HMP', 'EMPclosed', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA']
if ref == 'OpenRef': GoodNames = ['MGRAST', 'HMP', 'EMPopen', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA']
for name in os.listdir(mydir +'data/micro'):
if name in GoodNames: pass
else: continue
if Ones == 'N': path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y': path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'micro', num_lines])
print name, num_lines
for name in os.listdir(mydir +'data/macro'):
if name in GoodNames: pass
else: continue
if Ones == 'N': path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y': path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'macro', num_lines])
print name, num_lines
metrics = ['log-modulo skewness', 'log-skew']
fig = plt.figure()
for index, i in enumerate(metrics):
metric = i
fig.add_subplot(2, 2, index+1)
fs = 10 # font size used across figures
MicIntList, MicCoefList, MacIntList, MacCoefList, R2List, metlist = [[], [], [], [], [], []]
Nlist, Slist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList = [[], [], [], [], [], [], [], [], []]
EvarList, EQList, OList = [[],[],[]]
SkewList, LogSkewList = [[],[]]
its = 1000
for n in range(its):
#print n, metric
Nlist, Slist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList = [[], [], [], [], [], [], [], [], []]
EvarList, EQList, OList = [[],[],[]]
SkewList, LogSkewList = [[],[]]
numMac = 0
numMic = 0
radDATA = []
for dataset in datasets:
name, kind, numlines = dataset
lines = []
if name == 'EMPclosed' or name == 'EMPopen':
lines = np.random.choice(range(1, numlines+1), 100, replace=True)
elif kind == 'micro': lines = np.random.choice(range(1, numlines+1), 100, replace=True)
else: lines = np.random.choice(range(1, numlines+1), 60, replace=True)
if Ones == 'N': path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y': path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
KindList.append(kind)
N = float(N)
S = float(S)
if S < 10 or N < 10: continue # Min species richness
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
# Rarity
lms = np.log10(np.abs(float(skew)) + 1)
if skew < 0: lms = lms * -1
SkewList.append(float(lms))
LogSkewList.append(float(logskew))
if kind == 'micro':
numMic += 1
klist.append('b')
if kind == 'macro':
klist.append('r')
numMac += 1
if index == 0: metlist = list(SkewList)
elif index == 1: metlist = list(LogSkewList)
# Multiple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
f = smf.ols('y ~ N * Kind', d).fit()
MacIntList.append(f.params[0])
MacCoefList.append(f.params[2])
if f.pvalues[1] < 0.05:
MicIntList.append(f.params[1] + f.params[0])
else:
MicIntList.append(f.params[0])
if f.pvalues[3] < 0.05:
MicCoefList.append(f.params[3] + f.params[2])
else:
MicCoefList.append(f.params[2])
R2List.append(f.rsquared)
MacListX = []
MacListY = []
MicListX = []
MicListY = []
for j, k in enumerate(KindList):
if k == 'micro':
MicListX.append(Nlist[j])
MicListY.append(metlist[j])
elif k == 'macro':
MacListX.append(Nlist[j])
MacListY.append(metlist[j])
MacPIx, MacFitted, MicPIx, MicFitted = [[],[],[],[]]
macCiH, macCiL, micCiH, micCiL = [[],[],[],[]]
lm = smf.ols('y ~ N * Kind', d).fit()
#print metric, '\n', lm.summary()
#f1 = smf.ols('y ~ N', d).fit()
#print metric, '\n', f1.summary()
st, data, ss2 = summary_table(lm, alpha=0.05)
fittedvalues = data[:,2]
predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
for j, kval in enumerate(KindList):
if kval == 'macro':
macCiH.append(predict_mean_ci_upp[j])
macCiL.append(predict_mean_ci_low[j])
MacPIx.append(Nlist[j])
MacFitted.append(f.fittedvalues[j])
elif kval == 'micro':
micCiH.append(predict_mean_ci_upp[j])
micCiL.append(predict_mean_ci_low[j])
MicPIx.append(Nlist[j])
MicFitted.append(f.fittedvalues[j])
MicPIx, MicFitted, micCiH, micCiL = zip(*sorted(zip(MicPIx, MicFitted, micCiH, micCiL)))
MacPIx, MacFitted, macCiH, macCiL = zip(*sorted(zip(MacPIx, MacFitted, macCiH, macCiL)))
for i in range(len(MicListX)):
plt.scatter(MacListX[i], MacListY[i], color = 'LightCoral', alpha= 1 , s = 4, linewidths=0.5, edgecolor='Crimson')
plt.scatter(MicListX[i], MicListY[i], color = 'SkyBlue', alpha= 1 , s = 4, linewidths=0.5, edgecolor='Steelblue')
plt.fill_between(MacPIx, macCiL, macCiH, color='r', lw=0.0, alpha=0.3)
plt.fill_between(MicPIx, micCiL, micCiH, color='b', lw=0.0, alpha=0.3)
MicInt = round(np.mean(MicIntList), 2)
MicCoef = round(np.mean(MicCoefList), 2)
MacInt = round(np.mean(MacIntList), 2)
MacCoef = round(np.mean(MacCoefList), 2)
r2 = round(np.mean(R2List), 2)
if index == 0:
plt.ylim(0, 2.5)
plt.xlim(0, 7)
plt.text(0.3, 2.2, r'$micro$'+ ' = '+str(round(MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs-1, color='Steelblue')
plt.text(0.3, 2.0, r'$macro$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs-1, color='Crimson')
plt.text(0.3, 1.7, r'$R^2$' + '=' +str(round(r2,3)), fontsize=fs-1, color='k')
if index == 1:
plt.ylim(-1, 4.5)
plt.xlim(0, 7)
plt.text(0.3, 4.0, r'$micro$'+ ' = '+str(round(MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs-1, color='Steelblue')
plt.text(0.3, 3.5, r'$macro$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs-1, color='Crimson')
plt.text(0.3, 2.9, r'$R^2$' + '=' +str(round(r2,3)), fontsize=fs-1, color='k')
plt.xlabel('Number of reads or individuals, '+ '$log$'+r'$_{10}$', fontsize=fs)
plt.ylabel(metric, fontsize=fs)
plt.tick_params(axis='both', which='major', labelsize=fs-1)
plt.subplots_adjust(wspace=0.4, hspace=0.4)
if ref == 'OpenRef'and Ones =='N': plt.savefig(mydir+'/figs/appendix/Rarity/SupplementaryRarityFig-OpenRef_NoMicrobe1s.png', dpi=600, bbox_inches = "tight")
elif ref == 'OpenRef'and Ones =='Y': plt.savefig(mydir+'/figs/appendix/Rarity/SupplementaryRarityFig-OpenRef.png', dpi=600, bbox_inches = "tight")
elif ref == 'ClosedRef'and Ones =='Y': plt.savefig(mydir+'/figs/appendix/Rarity/SupplementaryRarityFig-ClosedRef.png', dpi=600, bbox_inches = "tight")
elif ref == 'ClosedRef'and Ones =='N': plt.savefig(mydir+'/figs/appendix/Rarity/SupplementaryRarityFig-ClosedRef_NoMicrobe1s.png', dpi=600, bbox_inches = "tight")
#plt.show()
return
def plotErrDistBug(in_filename,gold_in_filename,out_filename,title):
errs = []
abs_errs = []
mses = []
# read data
if in_filename == None:
abs_errs = None
mses = None
else:
in_file = open(in_filename,'rb')
errs = cPickle.load(in_file)
abs_errs = cPickle.load(in_file)
mses = cPickle.load(in_file)
in_file.close()
g_in_file = open(gold_in_filename,'rb')
g_errs = cPickle.load(g_in_file)
g_abs_errs = cPickle.load(g_in_file)
g_mses = cPickle.load(g_in_file)
g_in_file.close()
# ensure that we don't actually take the log of 0
g_mses = np.array(g_mses)
g_abs_errs = np.array(g_abs_errs)
zeros = g_mses>0.0
g_mses = g_mses[zeros]
g_abs_errs = g_abs_errs[zeros]
#zeros = g_abs_errs>0.0
zeros = g_abs_errs>1.0E-10
g_mses = g_mses[zeros]
g_abs_errs = g_abs_errs[zeros]
#print g_mses
#print g_abs_errs
#print g_mses
if not mses == None:
mses = np.array(mses)
abs_errs = np.array(abs_errs)
zeros = mses>0.0
mses = mses[zeros]
abs_errs = abs_errs[zeros]
#zeros = abs_errs>0.0
zeros = abs_errs>1.0E-10
mses = mses[zeros]
abs_errs = abs_errs[zeros]
print 'mses: ' + `mses`
#print g_mses
#print np.log2(g_mses)
g_dist = np.divide(g_mses,g_abs_errs)
if not mses == None:
dist = np.divide(mses,abs_errs)
# determine Ordinary Least Squares
X = np.log2(g_abs_errs)
X = sm.add_constant(X)
#print X
#print len(g_abs_errs)
#model = sm.OLS(g_mses,X)
model = sm.OLS(np.log2(g_mses),X)
#model = sm.RLM(np.log2(g_mses),X)
results = model.fit()
#print results.params
#print results.summary()
#print dir(results)
#print results.outlier_test()
prstd, iv_l, iv_u = wls_prediction_std(results)
st, data, ss2 = summary_table(results, alpha=0.05)
#print oi.OLSInfluence(results).influence
fittedvalues = data[:,2]
predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
# check we got the right things
#print np.max(np.abs(results.fittedvalues - fittedvalues))
#print np.max(np.abs(iv_l - predict_ci_low))
#print np.max(np.abs(iv_u - predict_ci_upp))
if not in_filename == None:
in_files_pattern = re.compile('\D*(\d+)\D*')
match = in_files_pattern.search(in_filename)
mutation_num = match.group(1)
module_name = re.sub('Gold','',gold_in_filename)
module_name = re.sub('(.*/)*','',module_name)
fig = plt.figure(figsize=(10.0, 7.0))
legend = []
if not mses == None:
legend.append('Mutation ' + `mutation_num`)
#plt.loglog(abs_errs, mses, basex=2, linestyle='', marker='+', color='b')
plt.plot(np.log2(abs_errs), np.log2(mses), linestyle='', marker='+', color='b')
legend.append('Mutation ' + `mutation_num` + ' -- outliers')
o_x,o_y = returnOutliers(results,sm.add_constant(np.log2(abs_errs)),np.log2(mses),alpha=0.05)
plt.plot(o_x, o_y, linestyle='', marker='*', color='b')
#plt.plot(abs_errs, dist, linestyle='', marker='o', color='b')
print 'Mutation outliers: ' + `len(o_x)`
legend.append('Bug-free')
#plt.loglog(g_abs_errs, g_mses, basex=2, linestyle='', marker='+', color='r')
plt.plot(np.log2(g_abs_errs), np.log2(g_mses), linestyle='', marker='+', color='r')
g_o_x,g_o_y = returnOutliers(results,sm.add_constant(np.log2(g_abs_errs)),np.log2(g_mses),alpha=0.05)
legend.append('Bug-free -- outliers')
plt.plot(g_o_x, g_o_y, linestyle='', marker='*', color='r')
#legend.append('Bug-free -- MSE/ABS')
#plt.plot(g_abs_errs, g_dist, linestyle='', marker='o', color='r')
print 'Gold outliers: ' + `len(g_o_x)`
legend.append('95% CI -')
#plt.loglog(g_abs_errs, iv_l, basex=2, linestyle='-', color='c')
plt.plot(np.log2(g_abs_errs), iv_l, linestyle='-', color='c')
#legend.append('95% CI - manual')
#plt.plot(np.log2(g_abs_errs), predict_ci_low, linestyle='-', color='m')
#plt.loglog(g_abs_errs, results.fittedvalues, basex=2, linestyle='-', color='k')
plt.plot(np.log2(g_abs_errs), results.fittedvalues, linestyle='-', color='k')
legend.append('95% CI +')
#plt.loglog(g_abs_errs, iv_u, basex=2, linestyle='-', color='g')
plt.plot(np.log2(g_abs_errs), iv_u, linestyle='-', color='g')
#legend.append('95% CI + manual')
#plt.plot(np.log2(g_abs_errs), predict_ci_upp, linestyle='-', color='y')
leg = plt.legend(legend, 'lower right',ncol=1)
# fix up plotting to look nice
plt.suptitle(module_name+' '+title, fontsize=35)
plt.xlabel('Log2(Execution Error)', fontsize=23)
plt.ylabel('Log2(Output Error)', fontsize=23)
if not out_filename == None:
plt.gcf().savefig(out_filename)
plt.show()
y = np.concatenate([y[0],y[1],y[2],y[3],y[4],y[5]])
x = np.concatenate([x[0],x[1],x[2],x[3],x[4],x[5]])
xsort = np.argsort(x)
x = x[xsort]
y = y[xsort]
xb = sm.add_constant(x)
model = sm.OLS(y,xb)
results = model.fit()
x2 = np.linspace(np.min(x),np.max(x),np.size(x))
y2 = results.predict(sm.add_constant(x2))
st, data, ss2 = summary_table(results, alpha=0.001)
fittedvalues = data[:,2]
predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
plt.scatter(x,y)
#plt.plot(x2,y2,'r')
plt.plot(x, fittedvalues, 'k', lw=2)
#plt.plot(x, predict_ci_low, 'r--', lw=2)
#plt.plot(x, predict_ci_upp, 'r--', lw=2)
plt.plot(x, predict_mean_ci_low, 'r--', lw=2)
plt.plot(x, predict_mean_ci_upp, 'r--', lw=2)
plt.xlabel('Max stream function at 26N (Sv)')
plt.ylabel('AMO-box SST anomaly')
plt.savefig('/home/ph290/Documents/figures/amoc_v_sst.png')
def modelcomparison():
OUT = open(mydir + 'output/model_comparison.txt','w+')
datasets = []
GoodNames = ['empclosed', 'HMP', 'BIGN', 'TARA', 'BOVINE', 'HUMAN', 'LAUB', 'SED', 'CHU', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'] # all microbe data is MGRAST
for name in os.listdir(mydir +'data/micro'):
if name in GoodNames: pass
else: continue
#path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'micro', num_lines])
print>>OUT, name, num_lines
for name in os.listdir(mydir +'data/macro'):
if name in GoodNames: pass
else: continue
#path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'macro', num_lines])
print>>OUT, name, num_lines
rarity = []
dominance = []
evenness = []
richness = []
Nlist = []
metrics = ['Rarity', 'Dominance', 'Evenness', 'Richness']
for index, i in enumerate(metrics):
print i, ': R-squared : AIC : BIC'
print>>OUT, i, ': R-squared : AIC : BIC'
loglogR2s, linlogR2s, linearR2s, loglinR2s = [[],[],[],[]]
loglogAICs, linlogAICs, linearAICs, loglinAICs = [[],[],[],[]]
loglogBICs, linlogBICs, linearBICs, loglinBICs = [[],[],[],[]]
its = 10
for n in range(its):
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
radDATA = []
for dataset in datasets:
name, kind, numlines = dataset
lines = []
if name == 'EMPclosed' or name == 'EMPopen':
lines = np.random.choice(range(1, numlines+1), 100, replace=True)
elif kind == 'micro': lines = np.random.choice(range(1, numlines+1), 100, replace=True)
else: lines = np.random.choice(range(1, numlines+1), 60, replace=True)
#path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
N = float(N)
S = float(S)
if S < 2 or N < 10: continue
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
ESimplist.append(float(np.log10(float(ESimp))))
KindList.append(kind)
BPlist.append(float(BP))
NmaxList.append(float(np.log10(float(Nmax))))
# log-modulo transformation of skewnness
lms = np.log10(np.abs(float(skew)) + 1)
if skew < 0: lms = lms * -1
rareSkews.append(float(lms))
if index == 0: metlist = list(rareSkews)
elif index == 1: metlist = list(NmaxList)
elif index == 2: metlist = list(ESimplist)
elif index == 3: metlist = list(Slist)
# Multiple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
loglog = smf.ols('y ~ N * Kind', d).fit()
loglogR2s.append(loglog.rsquared)
loglogAICs.append(loglog.aic)
loglogBICs.append(loglog.bic)
# Multiple regression
xlist = 10**np.array(Nlist)
d = pd.DataFrame({'N': list(xlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
loglin = smf.ols('y ~ N * Kind', d).fit()
loglinR2s.append(loglin.rsquared)
loglinAICs.append(loglin.aic)
loglinBICs.append(loglin.bic)
# Multiple regression
ylist = 10**np.array(metlist)
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(ylist)
d['Kind'] = list(KindList)
linlog = smf.ols('y ~ N * Kind', d).fit()
linlogR2s.append(linlog.rsquared)
linlogAICs.append(linlog.aic)
linlogBICs.append(linlog.bic)
# Multiple regression
ylist = 10**np.array(metlist)
xlist = 10**np.array(Nlist)
d = pd.DataFrame({'N': list(xlist)})
d['y'] = list(ylist)
d['Kind'] = list(KindList)
linear = smf.ols('y ~ N * Kind', d).fit()
linearR2s.append(linear.rsquared)
linearAICs.append(linear.aic)
linearBICs.append(linear.bic)
st, data, ss2 = summary_table(linear, alpha=0.05)
#fittedvalues = data[:,2]
#predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
avgloglogR2 = round(np.mean(loglogR2s),3)
avglinlogR2 = round(np.mean(linlogR2s),3)
avglinearR2 = round(np.mean(linearR2s),3)
avgloglinR2 = round(np.mean(loglinR2s),3)
avgloglogAIC = round(np.mean(loglogAICs),3)
avglinlogAIC = round(np.mean(linlogAICs),3)
avglinearAIC = round(np.mean(linearAICs),3)
avgloglinAIC = round(np.mean(loglinAICs),3)
avgloglogBIC = round(np.mean(loglogBICs),3)
avglinlogBIC = round(np.mean(linlogBICs),3)
avglinearBIC = round(np.mean(linearBICs),3)
avgloglinBIC = round(np.mean(loglinBICs),3)
print 'power-law: ', avgloglogR2,' ', avgloglogAIC,' ', avgloglogBIC
print>>OUT, 'averages from power-law', avgloglogR2,' ',avgloglogAIC,' ', avgloglogBIC
print 'semilog: ', avglinlogR2,' ', avglinlogAIC,' ', avglinlogBIC
print>>OUT,'averages from semilog', avglinlogR2,' ', avglinlogAIC,' ', avglinlogBIC
print 'exponential: ', avgloglinR2,' ', avgloglinAIC,' ', avgloglinBIC
print>>OUT,'averages from exponential', avgloglinR2,' ', avgloglinAIC,' ', avgloglinBIC
print 'linear: ', avglinearR2,' ', avglinearAIC,' ', avglinearBIC,'\n'
print>>OUT,'averages from linear', avglinearR2,' ', avglinearAIC,' ', avglinearBIC,'\n'
OUT.close()
return
def Fig1(condition, ones, sampling):
tail = str()
if ones is False:
tail = '-SADMetricData_NoMicrobe1s.txt'
elif ones is True:
tail = '-SADMetricData.txt'
datasets = []
GoodNames = []
emp = str()
if condition == 'open': emp = 'EMPopen'
elif condition == 'closed': emp = 'EMPclosed'
#GoodNames = [emp, 'HMP', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA']
#GoodNames = [emp, 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'] # all microbe data is emp
#GoodNames = ['HMP', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'] # all microbe data is HMP
GoodNames = [emp, 'HMP', 'BIGN', 'TARA', 'BOVINE', 'HUMAN', 'LAUB', 'SED', 'CHU', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'] # all microbe data is MGRAST
for name in os.listdir(mydir +'data/micro'):
if name in GoodNames: pass
else: continue
path = mydir+'data/micro/'+name+'/'+name+tail
num_lines = sum(1 for line in open(path))
datasets.append([name, 'micro', num_lines])
print name, num_lines
for name in os.listdir(mydir +'data/macro'):
if name in GoodNames: pass
else: continue
path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'macro', num_lines])
print name, num_lines
metrics = ['Rarity, '+r'$log_{10}$',
'Dominance, '+r'$log_{10}$',
'Evenness, ' +r'$log_{10}$',
'Richness, ' +r'$log_{10}$',] #+r'$(S)^{2}$']
fig = plt.figure()
for index, i in enumerate(metrics):
metric = i
fig.add_subplot(2, 2, index+1)
fs = 12 # font size used across figures
MicIntList, MicCoefList, MacIntList, MacCoefList, R2List, metlist = [[], [], [], [], [], []]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
its = 10
for n in range(its):
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
numMac = 0
numMic = 0
radDATA = []
for dataset in datasets:
name, kind, numlines = dataset
lines = []
small = ['BIGN', 'BOVINE', 'CHU', 'LAUB', 'SED']
big = ['HUMAN', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO']
if kind == 'macro':
lines = np.random.choice(range(1, numlines+1), 100, replace=True)
elif name in small:
lines = np.random.choice(range(1, numlines+1), 20, replace=True)
elif name in big:
lines = np.random.choice(range(1, numlines+1), 50, replace=True)
elif name == 'TARA':
lines = np.random.choice(range(1, numlines+1), 50, replace=True)
else:
lines = np.random.choice(range(1, numlines+1), 50, replace=True)
if kind == 'micro':
path = mydir+'data/'+kind+'/'+name+'/'+name+tail
else:
path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
N = float(N)
S = float(S)
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
ESimplist.append(float(np.log10(float(ESimp))))
KindList.append(kind)
BPlist.append(float(BP))
NmaxList.append(float(np.log10(float(Nmax))))
# log-modulo transformation of skewnness
lms = np.log10(np.abs(float(skew)) + 1)
if skew < 0: lms = lms * -1
rareSkews.append(float(lms))
if kind == 'micro':
numMic += 1
klist.append('b')
if kind == 'macro':
klist.append('r')
numMac += 1
if index == 0: metlist = list(rareSkews)
elif index == 1: metlist = list(NmaxList)
elif index == 2: metlist = list(ESimplist)
elif index == 3: metlist = list(Slist)
# Multiple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
f = smf.ols('y ~ N * Kind', d).fit()
#f = smf.rlm('y ~ N * Kind', d).fit()
#r2 = smf.wls('y ~ N * Kind', d, weights= f.weights).fit().rsquared
r2 = f.rsquared
MacIntList.append(f.params[0])
MacCoefList.append(f.params[2])
if f.pvalues[1] < 0.05:
MicIntList.append(f.params[1] + f.params[0])
else:
MicIntList.append(f.params[0])
if f.pvalues[3] < 0.05:
MicCoefList.append(f.params[3] + f.params[2])
else:
MicCoefList.append(f.params[2])
R2List.append(r2)
MacPIx, MacFitted, MicPIx, MicFitted = [[],[],[],[]]
macCiH, macCiL, micCiH, micCiL = [[],[],[],[]]
MacListX = []
MacListY = []
MicListX = []
MicListY = []
for j, k in enumerate(KindList):
if k == 'micro':
MicListX.append(Nlist[j])
MicListY.append(metlist[j])
elif k == 'macro':
MacListX.append(Nlist[j])
MacListY.append(metlist[j])
print metric
ols = smf.ols('y ~ N * Kind', d).fit()
#rlm = smf.rlm('y ~ N * Kind', d).fit()
#wls = smf.wls('y ~ N * Kind', d, weights= rlm.weights).fit()
#r2 = wls.rsquared
r2 = ols.rsquared
st, data, ss2 = summary_table(ols, alpha=0.05)
# ss2: Obs, Dep Var Population, Predicted Value, Std Error Mean Predict,
# Mean ci 95% low, Mean ci 95% upp, Predict ci 95% low, Predict ci 95% upp,
# Residual, Std Error Residual, Student Residual, Cook's D
#fittedvalues = data[:,2]
#predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
for j, kval in enumerate(KindList):
if kval == 'macro':
macCiH.append(predict_mean_ci_upp[j])
macCiL.append(predict_mean_ci_low[j])
MacPIx.append(Nlist[j])
MacFitted.append(ols.fittedvalues[j])
elif kval == 'micro':
micCiH.append(predict_mean_ci_upp[j])
micCiL.append(predict_mean_ci_low[j])
MicPIx.append(Nlist[j])
MicFitted.append(ols.fittedvalues[j])
MicPIx, MicFitted, micCiH, micCiL = zip(*sorted(zip(MicPIx, MicFitted, micCiH, micCiL)))
MacPIx, MacFitted, macCiH, macCiL = zip(*sorted(zip(MacPIx, MacFitted, macCiH, macCiL)))
num = min(len(MacListX), len(MicListX))
micnums = np.random.choice(range(0, len(MicListX)), num, replace=False)
macnums = np.random.choice(range(0, len(MacListX)), num, replace=False)
for i, ind in enumerate(micnums):
plt.scatter(MacListX[macnums[i]], MacListY[macnums[i]], color = 'LightCoral', alpha= 1 , s = 8, linewidths=0.5, edgecolor='Crimson')
plt.scatter(MicListX[ind], MicListY[ind], color = 'SkyBlue', alpha= 1 , s = 8, linewidths=0.5, edgecolor='Steelblue')
plt.fill_between(MacPIx, macCiL, macCiH, color='LightCoral', lw=0.0, alpha=0.9)
plt.plot(MacPIx, MacFitted, color='r', ls='--', lw=0.5, alpha=0.9)
plt.fill_between(MicPIx, micCiL, micCiH, color='b', lw=0.0, alpha=0.3)
plt.plot(MicPIx, MicFitted, color='b', ls='--', lw=0.5, alpha=0.9)
MicInt = round(np.mean(MicIntList), 2)
MicCoef = round(np.mean(MicCoefList), 2)
MacInt = round(np.mean(MacIntList), 2)
MacCoef = round(np.mean(MacCoefList), 2)
R2 = round(np.mean(R2List), 2)
if index == 0:
plt.ylim(-0.1, 2.5)
plt.xlim(0, 8.2)
plt.text(0.35, 2.1, r'$micro$'+ ' = '+str(round(10**MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(0.35, 1.8, r'$macro$'+ ' = '+str(round(10**MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(0.35, 1.4, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
plt.scatter([0],[-1], color = 'SkyBlue', alpha = 1, s=15, linewidths=0.9, edgecolor='Steelblue', label= 'microbes (n='+str(len(MicListY))+')')
plt.scatter([0],[-1], color = 'LightCoral',alpha= 1, s=15, linewidths=0.9, edgecolor='Crimson', label= 'macrobes (n='+str(len(MacListY))+')')
plt.legend(bbox_to_anchor=(-0.04, 1.05, 2.48, .2), loc=10, ncol=2, mode="expand",prop={'size':fs})
elif index == 1:
plt.plot([0,8.2],[0,8.2], ls = '--', lw=1, c='0.7')
plt.ylim(0, 8)
plt.xlim(0, 8.2)
plt.text(0.35, 6.7, r'$micro$'+ ' = '+str(round(10**MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(0.35, 5.7, r'$macro$'+ ' = '+str(round(10**MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(0.35, 4.7, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
elif index == 2:
plt.ylim(-3.5, 0.0)
plt.xlim(0, 8.2)
plt.text(0.35, -2.9, r'$micro$'+ ' = '+str(round(10**MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(0.35, -3.3, r'$macro$'+ ' = '+str(round(10**MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(0.35, -2.5, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
elif index == 3:
plt.ylim(0.9, 5.0)
plt.xlim(0, 8.2)
plt.text(0.35, 4.5, r'$micro$'+ ' = '+str(round(2**MicInt,2))+'*'+r'$N$'+'$^{'+str(round(MicCoef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(0.35, 4.0, r'$macro$'+ ' = '+str(round(2**MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(0.35, 3.5, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
print condition, ones, ': S =', '%.3e' % (10**(MicInt + MicCoef*(30.0)))
#print R2
plt.xlabel('$log$'+r'$_{10}$'+'($N$)', fontsize=fs)
plt.ylabel(metric, fontsize=fs)
plt.tick_params(axis='both', which='major', labelsize=fs-3)
plt.subplots_adjust(wspace=0.4, hspace=0.4)
if ones == False:
plt.savefig(mydir+'/figs/Fig1/Locey_Lennon_2015_Fig1-'+condition+'_NoSingletons_'+str(sampling)+'.pdf', dpi=300, bbox_inches = "tight")
if ones == True:
plt.savefig(mydir+'/figs/Fig1/Locey_Lennon_2015_Fig1-'+condition+'_'+str(sampling)+'.pdf', dpi=300, bbox_inches = "tight")
#plt.show()
plt.close()
return
if index == 0: metlist = list(Rs)
elif index == 1: metlist = list(Nmaxs)
elif index == 2: metlist = list(Evs)
elif index == 3: metlist = list(Ss)
print len(Ns), len(metlist)
d = pd.DataFrame({'N': list(Ns)})
d['y'] = list(metlist)
f = smf.ols('y ~ N', d).fit()
r2 = round(f.rsquared, 2)
Int = f.params[0]
Coef = f.params[1]
st, data, ss2 = summary_table(f, alpha=0.05)
# ss2: Obs, Dep Var Population, Predicted Value, Std Error Mean Predict,
# Mean ci 95% low, Mean ci 95% upp, Predict ci 95% low, Predict ci 95% upp,
# Residual, Std Error Residual, Student Residual, Cook's D
fitted = data[:, 2]
#predict_mean_se = data[:,3]
mean_ci_low, mean_ci_upp = data[:, 4:6].T
ci_low, ci_upp = data[:, 6:8].T
ci_Ns = data[:, 0]
Ns, metlist, fitted, ci_low, ci_upp = zip(
*sorted(zip(Ns, metlist, fitted, ci_low, ci_upp)))
plt.scatter(Ns,
metlist,
#### plot figure ###############################################################
xlab = r"$log_{10}$" + "(" + r"$\tau$" + ")"
fs = 8 # fontsize
fig = plt.figure()
#### N vs. Tau #################################################################
fig.add_subplot(2, 2, 1)
f2 = smf.ols("N ~ tau + I(tau ** 2.0)", d).fit()
print f2.summary()
a, b, c = f2.params
p1, p2, p3 = f2.pvalues
r2 = round(f2.rsquared, 2)
st, data, ss2 = summary_table(f2, alpha=0.05)
fitted = data[:, 2]
pred_mean_se = data[:, 3]
pred_mean_ci_low, pred_mean_ci_upp = data[:, 4:6].T
pred_ci_low, pred_ci_upp = data[:, 6:8].T
tau2, fitted, pred_ci_low, pred_ci_upp, pred_mean_ci_low, pred_mean_ci_upp = zip(
*sorted(zip(tau, fitted, pred_ci_low, pred_ci_upp, pred_mean_ci_low, pred_mean_ci_upp))
)
plt.scatter(tau, N, color=colors, s=10, linewidths=0.1, edgecolor="k")
# plt.fill_between(tau2, pred_ci_low, pred_ci_upp, color='r', lw=0.0, alpha=0.1)
# plt.fill_between(tau2, pred_mean_ci_low, pred_mean_ci_upp, color='r', lw=0.0, alpha=0.3)
plt.plot(tau2, fitted, color="r", ls="--", lw=1.5, alpha=0.9)
plt.ylabel(r"$log_{10}$" + "(" + r"$N$" + ")", fontsize=fs + 6)
def figplot(x, y, xlab, ylab, fig, n):
'''main figure plotting function'''
fig.add_subplot(2, 2, n)
y2 = list(y)
x2 = list(x)
d = pd.DataFrame({'x': list(x2)})
d['y'] = list(y2)
f = smf.ols('y ~ x', d).fit()
m, b, r, p, std_err = stats.linregress(x2, y2)
st, data, ss2 = summary_table(f, alpha=0.05)
fitted = data[:, 2]
mean_ci_low, mean_ci_upp = data[:, 4:6].T
ci_low, ci_upp = data[:, 6:8].T
x2, y2, fitted, ci_low, ci_upp = zip(
*sorted(zip(x2, y2, fitted, ci_low, ci_upp)))
if n == 1:
lab = r'$R_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$R_{microbes}$' + ' = 2.34*' + r'$N$' + '$^{0.14}$' + '\n'
lab += r'$R_{macrobes}$' + ' = 1.7*' + r'$N$' + '$^{0.11}$'
plt.text(0.2, 0.8, lab, fontsize=7)
elif n == 2:
lab = r'$D_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$D_{microbes}$' + ' = 0.44*' + r'$N$' + '$^{0.92}$' + '\n'
lab += r'$D_{macrobes}$' + ' = 0.23*' + r'$N$' + '$^{0.99}$'
plt.text(0.2, 3.0, lab, fontsize=7)
elif n == 3:
lab = r'$E_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$E_{microbes}$' + ' = 0.58*' + r'$N$' + '$^{-0.23}$' + '\n'
lab += r'$E_{macrobes}$' + ' = 1.15*' + r'$N$' + '$^{-0.21}$'
plt.text(0.2, -1.7, lab, fontsize=7)
elif n == 4:
lab = r'$S_{models}$' + ' = ' + str(round(
10**b, 2)) + '*' + r'$N$' + '$^{' + str(round(m, 2)) + '}$' + '\n'
lab += r'$S_{microbes}$' + ' = 1.77*' + r'$N$' + '$^{0.38}$' + '\n'
lab += r'$S_{macrobes}$' + ' = 1.77*' + r'$N$' + '$^{0.24}$'
plt.text(0.2, 1.9, lab, fontsize=7)
#plt.hexbin(x2, y2, mincnt=1, gridsize = 40, bins='log', cmap=plt.cm.jet)
plt.scatter(x2,
y2,
color='SkyBlue',
alpha=1,
s=12,
linewidths=0.5,
edgecolor='Steelblue')
if n == 3: plt.legend(loc='best', fontsize=6, frameon=False)
plt.plot(x2, fitted, color='k', ls='--', lw=1.0, alpha=0.9)
plt.xlabel(xlab, fontsize=8)
plt.ylabel(ylab, fontsize=8)
plt.tick_params(axis='both', labelsize=5)
plt.xlim(0, 1.05 * max(x2))
if n == 1: plt.ylim(0.0, 1.1)
elif n == 2: plt.ylim(0.0, 4.2)
elif n == 3: plt.ylim(-1.8, 0.05)
elif n == 4: plt.ylim(0.4, 2.5)
return fig
def Fig1():
datasets = []
GoodNames = ['MGRAST', 'HMP', 'EMPclosed', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA']
for name in os.listdir(mydir +'data/micro'):
if name in GoodNames: pass
else: continue
#if name in BadNames: continue
#else: pass
#path = mydir2+'data/micro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir2+'data/micro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'micro', num_lines])
print name, num_lines
for name in os.listdir(mydir2 +'data/macro'):
if name in GoodNames: pass
else: continue
#if name in BadNames: continue
#else: pass
#path = mydir2+'data/macro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir2+'data/macro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'macro', num_lines])
print name, num_lines
metrics = ['Rarity, '+r'$log_{10}$',
'Dominance, '+r'$log_{10}$',
'Evenness, ' +r'$log_{10}$',
'Richness, ' +r'$log_{10}$']
fig = plt.figure()
for index, i in enumerate(metrics):
metric = i
fig.add_subplot(2, 2, index+1)
fs = 10 # font size used across figures
MicIntList, MicCoefList, MacIntList, MacCoefList, R2List, metlist = [[], [], [], [], [], []]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
its = 1
for n in range(its):
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
numMac = 0
numMic = 0
radDATA = []
for dataset in datasets:
name, kind, numlines = dataset
lines = []
if name == 'EMPclosed' or name == 'EMPopen':
lines = np.random.choice(range(1, numlines+1), 1000, replace=True) # 166
elif kind == 'micro': lines = np.random.choice(range(1, numlines+1), 1000, replace=True) #167
else: lines = np.random.choice(range(1, numlines+1), 600, replace=True) # 100
#path = mydir2+'data/'+kind+'/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir2+'data/'+kind+'/'+name+'/'+name+'-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
N = float(N)
S = float(S)
if S < 10 or N < 11: continue
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
ESimplist.append(float(np.log10(float(ESimp))))
kind = np.random.choice(['micro', 'macro'])
KindList.append(kind)
BPlist.append(float(BP))
NmaxList.append(float(np.log10(float(Nmax))))
# log-modulo transformation of skewnness
lms = np.log10(np.abs(float(skew)) + 1)
if skew < 0: lms = lms * -1
rareSkews.append(float(lms))
if kind == 'micro':
numMic += 1
klist.append('b')
if kind == 'macro':
klist.append('r')
numMac += 1
if index == 0: metlist = list(rareSkews)
elif index == 1: metlist = list(NmaxList)
elif index == 2: metlist = list(ESimplist)
elif index == 3: metlist = list(Slist)
# Multiple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
f = smf.ols('y ~ N * Kind', d).fit()
MacIntList.append(f.params[0])
MacCoefList.append(f.params[2])
if f.pvalues[1] < 0.05:
MicIntList.append(f.params[1] + f.params[0])
else:
MicIntList.append(f.params[0])
if f.pvalues[3] < 0.05:
MicCoefList.append(f.params[3] + f.params[2])
else:
MicCoefList.append(f.params[2])
R2List.append(f.rsquared)
MacPIx, MacFitted, MicPIx, MicFitted = [[],[],[],[]]
macCiH, macCiL, micCiH, micCiL = [[],[],[],[]]
MacListX = []
MacListY = []
MicListX = []
MicListY = []
for j, k in enumerate(KindList):
if k == 'micro':
MicListX.append(Nlist[j])
MicListY.append(metlist[j])
elif k == 'macro':
MacListX.append(Nlist[j])
MacListY.append(metlist[j])
print metric
lm = smf.ols('y ~ N * Kind', d).fit()
print lm.summary()
print '\n\n'
st, data, ss2 = summary_table(lm, alpha=0.05)
# ss2: Obs, Dep Var Population, Predicted Value, Std Error Mean Predict,
# Mean ci 95% low, Mean ci 95% upp, Predict ci 95% low, Predict ci 95% upp,
# Residual, Std Error Residual, Student Residual, Cook's D
fittedvalues = data[:,2]
predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
for j, kval in enumerate(KindList):
if kval == 'macro':
macCiH.append(predict_mean_ci_upp[j])
macCiL.append(predict_mean_ci_low[j])
MacPIx.append(Nlist[j])
MacFitted.append(f.fittedvalues[j])
elif kval == 'micro':
micCiH.append(predict_mean_ci_upp[j])
micCiL.append(predict_mean_ci_low[j])
MicPIx.append(Nlist[j])
MicFitted.append(f.fittedvalues[j])
MicPIx, MicFitted, micCiH, micCiL = zip(*sorted(zip(MicPIx, MicFitted, micCiH, micCiL)))
MacPIx, MacFitted, macCiH, macCiL = zip(*sorted(zip(MacPIx, MacFitted, macCiH, macCiL)))
num = min(len(MacListX), len(MicListX))
for i in range(num):
plt.scatter(MacListX[i], MacListY[i], color = '0.4', alpha= 1 , s = 4, linewidths=0.5, edgecolor='0.3')
plt.scatter(MicListX[i], MicListY[i], color = '0.4', alpha= 1 , s = 4, linewidths=0.5, edgecolor='0.3')
plt.fill_between(MacPIx, macCiL, macCiH, color='lime', lw=0.0, alpha=0.3)
plt.plot(MacPIx, MacFitted, color='lime', ls='--', lw=0.5, alpha=0.8)
MacInt = round(np.mean(MacIntList), 2)
MacCoef = round(np.mean(MacCoefList), 2)
R2 = round(np.mean(R2List), 2)
if index == 0:
plt.ylim(-0.1, 2.0)
plt.xlim(1, 7)
plt.text(1.35, 1.5, r'$Rarity$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='k')
plt.text(1.35, 1.2, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='0.3')
plt.scatter([0],[-1], color = 'SkyBlue', alpha = 1, s=10, linewidths=0.9, edgecolor='Steelblue', label= 'microbes (n='+str(len(MicListY))+')')
plt.scatter([0],[-1], color = 'LightCoral',alpha= 1, s=10, linewidths=0.9, edgecolor='Crimson', label= 'macrobes (n='+str(len(MacListY))+')')
plt.legend(bbox_to_anchor=(-0.04, 1.1, 2.48, .2), loc=10, ncol=2, mode="expand",prop={'size':fs+2})
elif index == 1:
plt.plot([0,7],[0,7], ls = '--', lw=1, c='0.7')
#ax.text(18, 21, '1:1 line', fontsize=fs*1.0, rotation=40, color='0.7')
plt.ylim(0, 6)
plt.xlim(1, 7)
plt.text(1.35, 4.5, r'$Dominance$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='k')
plt.text(1.35, 3.75, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='0.3')
elif index == 2:
plt.ylim(-3.0, 0.0)
plt.xlim(0, 7)
plt.text(0.35, -2.5, r'$Evenness$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='k')
plt.text(0.35, -2.2, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='0.3')
elif index == 3:
plt.ylim(0.9, 4.5)
plt.xlim(1, 7)
plt.text(1.35, 3.5, r'$Richness$'+ ' = '+str(round(MacInt,2))+'*'+r'$N$'+'$^{'+str(round(MacCoef,2))+'}$', fontsize=fs, color='k')
plt.text(1.35, 3.0, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='0.3')
plt.xlabel('Number of reads or individuals, '+ '$log$'+r'$_{10}$', fontsize=fs)
plt.ylabel(metric, fontsize=fs)
plt.tick_params(axis='both', which='major', labelsize=fs-3)
plt.subplots_adjust(wspace=0.4, hspace=0.4)
#plt.savefig(mydir+'/figs/appendix/Fig1/RandomAssign/Locey_Lennon_2015_Pooled-OpenReference_NoSingletons.png', dpi=600, bbox_inches = "tight")
#plt.savefig(mydir+'/figs/appendix/Fig1/RandomAssign/Locey_Lennon_2015_Pooled-ClosedReference_NoSingletons.png', dpi=600, bbox_inches = "tight")
#plt.savefig(mydir+'/figs/appendix/Fig1/RandomAssign/Locey_Lennon_2015_Pooled-OpenReference.png', dpi=600, bbox_inches = "tight")
plt.savefig(mydir+'/figs/appendix/Fig1/RandomAssign/Locey_Lennon_2015_Pooled-ClosedReference.png', dpi=600, bbox_inches = "tight")
#plt.show()
#plt.close()
return
def d(re,alpha): st, data, ss2 = summary_table(re, alpha) return st,data,ss2
def Fig3(condition, ones, sampling):
""" A figure demonstrating a strong richness relationship across 10 or 11
orders of magnitude in total abundance. Taxonomic richness of a sample
scales in a log-log fashion with the total abundance of the sample.
"""
fs = 12 # font size used across figures
metric = 'Richness, '+r'$log$'+r'$_{10}$'
tail = str()
if ones is False:
tail = '-SADMetricData_NoMicrobe1s.txt'
elif ones is True:
tail = '-SADMetricData.txt'
datasets = []
GoodNames = []
emp = str()
if condition == 'open': emp = 'EMPopen'
elif condition == 'closed': emp = 'EMPclosed'
GoodNames = [emp, 'TARA', 'HMP', 'BIGN', 'BOVINE', 'CHU', 'LAUB', 'SED', 'HUMAN', 'CHINA', 'CATLIN', 'FUNGI']
print '\n'
its = 1
d_blist = []
d_zlist = []
s_blist = []
s_zlist = []
for name in os.listdir(mydir +'data/micro'):
if name in GoodNames: pass
else: continue
path = mydir+'data/micro/'+name+'/'+name+tail
numlines = sum(1 for line in open(path))
#print name, numlines
datasets.append([name, 'micro', numlines])
if sampling <= 500: its = 100
else: its = 100
for i in range(its):
Nlist, Slist, klist, NmaxList = [[],[],[],[]]
for dataset in datasets:
radDATA = []
name, kind, numlines = dataset
lines = []
small_mgrast = ['BIGN', 'BOVINE', 'CHU', 'LAUB', 'SED']
big_mgrast = ['HUMAN', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO']
if kind == 'micro':
if name in small_mgrast:
lines = np.random.choice(range(1, numlines+1), 160, replace=True) # 40
elif name in big_mgrast:
lines = np.random.choice(range(1, numlines+1), 400, replace=True) # 100
else:
lines = np.random.choice(range(1, numlines+1), 400, replace=True) # 100
path = mydir+'data/micro/'+name+'/'+name+tail
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
ct = 0
for data in radDATA:
data = data.split()
if data == []: continue
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
N = float(N)
S = float(S)
Nmax = float(Nmax)
ct += 1
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
NmaxList.append(float(np.log10(Nmax)))
klist.append('DarkCyan')
#print name, ct
Nlist, Slist, NmaxList = zip(*sorted(zip(Nlist, Slist, NmaxList)))
Nlist = list(Nlist)
Slist = list(Slist)
NmaxList = list(NmaxList)
# Regression for Dominance (Nmax) vs. N
d = pd.DataFrame({'N': Nlist})
d['Nmax'] = NmaxList
f = smf.ols('Nmax ~ N', d).fit()
R2 = f.rsquared
pval = f.pvalues[0]
intercept = f.params[0]
slope = f.params[1]
d_blist.append(intercept)
d_zlist.append(slope)
# Regression for Richness (S) vs. N
d = pd.DataFrame({'N': Nlist})
d['S'] = Slist
f = smf.ols('S ~ N', d).fit()
R2 = f.rsquared
pval = f.pvalues[0]
intercept = f.params[0]
slope = f.params[1]
s_blist.append(intercept)
s_zlist.append(slope)
sb = np.mean(s_blist)
sz = np.mean(s_zlist)
db = np.mean(d_blist)
dz = np.mean(d_zlist)
#print 'R2 for Nmax vs. N:', round(dR2, 3)
print 'Nmax =', round(10**db, 2), '*', 'N^', round(dz, 2)
#print 'R2 for S vs. N:', round(R2, 3)
print 'S =', round(10**sb, 2), '*', 'N^', round(sz, 2),'\n'
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
plt.text(2, 11.0, r'$S$'+ ' = '+str(round(10**sb, 1))+'*'+r'$N$'+'$^{'+str(round(sz, 2))+'},$' + r' $r^2$' + '=' +str(round(R2,2)), fontsize=fs+5, color='Crimson', alpha=0.9)
# code for prediction intervals
X = np.linspace(5, 32, 100)
Y = f.predict(exog=dict(N=X))
Nlist2 = Nlist + X.tolist()
Slist2 = Slist + Y.tolist()
d = pd.DataFrame({'N': list(Nlist2)})
d['y'] = list(Slist2)
f = smf.ols('y ~ N', d).fit()
st, data, ss2 = summary_table(f, alpha=0.05)
#fittedvalues = data[:,2]
#pred_mean_se = data[:,3]
pred_mean_ci_low, pred_mean_ci_upp = data[:,4:6].T
pred_ci_low, pred_ci_upp = data[:,6:8].T
plt.fill_between(Nlist2, pred_ci_low, pred_ci_upp, color='r', lw=0.5, alpha=0.2)
z = np.polyfit(Nlist2, Slist2, 1)
p = np.poly1d(z)
xp = np.linspace(0, 32, 1000)
label1 = 'Richness-abundance scaling relationship, $S$ = 7.6$N^{0.35}$'
label2 = 'Predicted $S$ via lognormal, published $N$ and $N_{max}$'
label3 = 'Predicted $S$ via lognormal, published $N$, $N_{max}$ = 0.4$N^{0.93}$'
plt.plot(xp, p(xp), '--', c='red', lw=2, alpha=0.8, color='Crimson', label=label1)
plt.scatter(Nlist, Slist, color = 'LightCoral', alpha= 1 , s = 10, linewidths=0.5, edgecolor='Crimson')
#plt.hexbin(Nlist, Slist, mincnt=1, gridsize = 80, bins='log', cmap=plt.cm.Reds_r, label='EMP')
# Adding in derived/inferred points
c = '0.3'
GO = [3.6*(10**28), 10.1*(10**28)] # estimated open ocean bacteria; Whitman et al. 1998
Pm = [2.8*(10**27), 3.0*(10**27)] # estimated Prochlorococcus; Flombaum et al. 2013
Syn = [6.7*(10**26), 7.3*(10**26)] # estimated Synechococcus; Flombaum et al. 2013
Earth = [9.2*(10**29), 31.7*(10**29)] # estimated bacteria on Earth; Kallmeyer et al. 2012
SAR11 = [2.0*(10**28), 2.0*(10**28)] # estimated percent abundance of SAR11; Morris et al. (2002)
HGx = [0.5*(10**14), 1.5*(10**14)] # estimated bacteria in Human gut; Berg (1996)
HGy = [0.05*min(HGx), 0.15*max(HGx)] # estimated most abundant bacteria in Human gut; Turnbaugh et al. (2009), & Dethlefsen et al. (2008)
COWx = [0.5*2.226*(10**15), 1.5*2.226*(10**15)] # estimated bacteria in Cow rumen; LOW: HIGH: Whitman et al. (1998)
COWy = [0.09*min(COWx), 0.15*max(COWx)] # estimated dominance in Cow rumen; Stevenson and Weimer (2006)
## PREDICTIONS OF S BASED ON THE EMPIRICAL S VS. N SCALING LAW, AND BASED
## ON THE LOGNORMAL PREDICTIVE FRAMEWORK OF CURTIS AND SLOAN USING
## 1.) THE ESTIMATED NMAX AND 2.) THE PREDICTED NMAX
Ns = []
Ss = []
DomSs = []
# Global Ocean estimates based on Whitman et al. (1998) and P. marinus (2012 paper)
guess = 0.1019
yrange = [min(Syn), max(Pm)]
Slist_ln, Slist_SvN, Dlist, Nlist = getS(GO, sb, sz, db, dz, guess, yrange, predictNmax=False)
S_ln = np.mean(Slist_ln)
S1 = float(S_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
Ss.append(S_ln)
print 'scaling law prediction of S for Global Ocean:', '%.3e' % 10**(S_SvN)
print 'lognormal prediction of S for Global Ocean, using estimated Nmax:', '%.3e' % 10**S_ln
guess = 0.1019
Slist_ln, Slist_SvN, Dlist, Nlist = getS(GO, sb, sz, db, dz, guess, yrange, predictNmax=True)
S_ln = np.mean(Slist_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
print 'lognormal prediction of S for Global Ocean, using predicted Nmax:', '%.3e' % 10**S_ln
#print 'P.m.:', '%.2e' % float(2.9*10**27), 'Nmax:', '%.2e' % 10**Nmax,'\n'
S2 = float(S_ln)
N = float(avgN)
S_sem = float(4*S_ln_sem)
N_sem = float(4*avgN_sem)
ax.text(13.5, S1*0.93, 'Global Ocean', fontsize=fs+2, color = 'k')
ax.axhline(S1, 0, 0.91, ls = '--', c = '0.6')
ax.text(N-1, S2*.80, 'Global ocean', fontsize=fs+2, color = 'k', rotation = 90)
ax.axvline(N, 0, 0.65, ls = '--', c = '0.6')
#plt.scatter([N], [S2], color = '0.2', alpha= 1 , s = 60, linewidths=1, edgecolor='k')
Ns.append(N)
DomSs.append(S2)
#plt.errorbar([N], [S2], xerr=N_sem, yerr=S_sem, color='k', linewidth=2)
# Earth, i.e., Global estimates based on Kallmeyer et al. (2012) and SAR11 (2002 paper)
guess = 0.1060
yrange = [min(Pm), max(SAR11)]
Slist_ln, Slist_SvN, Dlist, Nlist = getS(Earth, sb, sz, db, dz, guess, yrange, predictNmax=False)
S_ln = np.mean(Slist_ln)
S1 = float(S_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
Ss.append(S_ln)
#print 'average N and sem:' '%.3e' % 10**avgN, '%.3e' % 10**avgN_sem
#print 'average Nmax and sem:' '%.3e' % 10**Nmax, '%.3e' % 10**Nmax_sem
print '\nscaling law prediction of S for Earth:', '%.3e' % 10**S_SvN #,'%.3e' % 10**S_SvN_sem #, '%.3e' % S_SvN_CI
print 'lognormal prediction of S for Earth, using estimated Nmax:', '%.3e' % 10**S_ln #, '%.3e' % 10**S_ln_sem#, '%.3e' % S_ln_CI
guess = 0.1060
Slist_ln, Slist_SvN, Dlist, Nlist = getS(Earth, sb, sz, db, dz, guess, yrange, predictNmax=True)
S_ln = np.mean(Slist_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
print 'lognormal prediction of S for Earth, using predicted Nmax:', '%.3e' % 10**S_ln #, '%.3e' % 10**S_ln_sem#, '%.3e' % S_ln_CI
#print 'SAR11:', '%.2e' % float(2.4*10**28), 'Nmax:', '%.2e' % Nmax,'\n'
S2 = float(S_ln)
N = float(avgN)
S_sem = float(4*S_ln_sem)
N_sem = float(4*avgN_sem)
ax.text(25, S2*1.025, 'Earth', fontsize=fs+2, color = 'k')
ax.axhline(S2, 0, 0.95, ls = '--', c = '0.6')
ax.text(N-1, 8, 'Earth', fontsize=fs+2, color = 'k', rotation = 90)
ax.axvline(N, 0, 0.82, ls = '--', c = '0.6')
#plt.scatter([N], [S2], color = '0.2', alpha= 1 , s = 60, linewidths=1, edgecolor='k')
#plt.errorbar([N], [S2], xerr=N_sem, yerr=S_sem, color='k', linewidth=2)
Ns.append(N)
DomSs.append(S2)
# Human Gut
guess = 0.1509
Slist_ln, Slist_SvN, Dlist, Nlist = getS(HGx, sb, sz, db, dz, guess, HGy, predictNmax=False)
S_ln = np.mean(Slist_ln)
S1 = float(S_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
Ss.append(S_ln)
Slist_ln, Slist_SvN, Dlist, Nlist = getS(HGx, sb, sz, db, dz, guess, HGy, predictNmax=True)
S_ln = np.mean(Slist_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
S2 = float(S_ln)
N = float(avgN)
S_sem = float(4*S_ln_sem)
N_sem = float(4*avgN_sem)
ax.text(3.5, S2*.9, 'Human Gut', fontsize=fs+2, color = 'k')
ax.axhline(S2, 0, 0.41, ls = '--', c = '0.6')
ax.text(N-1, 3.6, 'Human Gut', fontsize=fs+2, color = 'k', rotation = 90)
ax.axvline(N, 0, 0.33, ls = '--', c = '0.6')
#plt.scatter([N], [S2], color = '0.2', alpha= 1 , s = 60, linewidths=1, edgecolor='k')
#plt.errorbar([N], [S2], xerr=N_sem, yerr=S_sem, color='k', linewidth=2)
Ns.append(N)
DomSs.append(S2)
#print 'predS for Human Gut:', '%.3e' % 10**S2
# Cow Rumen
guess = 0.1
Slist_ln, Slist_SvN, Dlist, Nlist = getS(COWx, sb, sz, db, dz, guess, COWy, predictNmax=False)
S_ln = np.mean(Slist_ln)
S1 = float(S_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
Ss.append(S_ln)
Slist_ln, Slist_SvN, Dlist, Nlist = getS(COWx, sb, sz, db, dz, guess, COWy, predictNmax=True)
S_ln = np.mean(Slist_ln)
S_ln_sem = stats.sem(Slist_ln, ddof=1)
S_SvN = np.mean(Slist_SvN)
S_SvN_sem = stats.sem(Slist_SvN, ddof=1)
Nmax = np.mean(Dlist)
Nmax_sem = stats.sem(Dlist, ddof=1)
avgN = np.mean(Nlist)
avgN_sem = stats.sem(Nlist, ddof=1)
S2 = float(S_ln)
N = float(avgN)
S_sem = float(4*S_ln_sem)
N_sem = float(4*avgN_sem)
ax.text(6, S2*1.04, 'Cow Rumen', fontsize=fs+2, color = 'k')
ax.axhline(S2, 0, 0.46, ls = '--', c = '0.6')
ax.text(N+0.3, 4.2, 'Cow Rumen', fontsize=fs+2, color = 'k', rotation = 90)
ax.axvline(N, 0, 0.38, ls = '--', c = '0.6')
Ns.append(N)
DomSs.append(S2)
plt.scatter(Ns, Ss, color = '0.4', alpha= 1, s = 50, linewidths=2, edgecolor='k', label=label2)
plt.scatter(Ns, DomSs, color = 'SkyBlue', alpha= 1, s = 50, linewidths=2, edgecolor='Steelblue', label=label3)
#plt.errorbar([N], [S2], xerr=N_sem, yerr=S_sem, color='k', linewidth=2)
ax.text(5, -0.8, 'Number of reads or total abundance, '+ '$log$'+r'$_{10}$', fontsize=fs+4)
ax.text(-2.1, 10, 'OTU '+ metric, fontsize=fs+4, rotation=90)
plt.xlim(1, 31)
plt.ylim(0.8, 14)
plt.legend(bbox_to_anchor=(-0.015, 1.03, 1.03, .2), loc=10, ncol=1,
mode="expand",prop={'size':fs+2.2})
if ones == False:
plt.savefig(mydir+'/figs/Fig3/figure3.pdf', dpi=300, bbox_inches = "tight")
#plt.savefig(mydir+'/figs/Fig3/Locey_Lennon_2015_Fig3-'+condition+'_NoSingletons_'+str(sampling)+'.pdf', dpi=300, bbox_inches = "tight")
if ones == True:
plt.savefig(mydir+'/figs/Fig3/figure3.pdf', dpi=300, bbox_inches = "tight")
#plt.savefig(mydir+'/figs/Fig3/Locey_Lennon_2015_Fig3-'+condition+'_'+str(sampling)+'.pdf', dpi=300, bbox_inches = "tight")
#plt.show()
return
def plot_time_series(fname_template=None,
region_number="all",
dpi=150,
ax = None,
melt_index_or_extent="index",
extent_melt_days_threshold=2,
include_ylabel=True,
gap_filled=True,
include_trendline=False,
include_trendline_only_if_significant=True,
include_legend_if_significant=True,
include_name_in_title=True,
print_trendline_summary=True,
offset_years_by_one=True,
add_confidence_intervals=True,
add_prediction_intervals=True,
verbose=True):
"""Cretae a plot of the time series of melt.
In fname_template, if you specify a {0} tag in the name, it will be filled
in with the region number. This is useful if you want to use region_number="all",
as that will create 8 plots for regions 0-7.
Use 'ax' to provide an axis upon which to draw. This is useful for putting
together a multi-part figure. Don't use this option if using "region_number="all",
as that will draw multiple plots on the same axes.
"""
if region_number == "all":
region_nums = range(8)
else:
region_nums = [region_number]
ax_provided = ax
for region_n in region_nums:
years, melt = get_time_series_data(region_number=region_n,
melt_index_or_extent=melt_index_or_extent,
extent_melt_days_threshold = extent_melt_days_threshold,
gap_filled=gap_filled,
return_in_km2=True)
# Since the "2019" melt season (.e.g) in Antarctica actually spans 2019-2020,
# it makes more sense to center it over the Jan 1, 2020 date rather than
# the start of 2019.
# Make it so.
if offset_years_by_one:
years = years + 1
if include_ylabel:
if max(melt) > 1e6:
melt = melt / 1e6
figure_exp = 6
else:
melt = melt / 1e3
figure_exp = 3
else:
melt = melt / 1e6
figure_exp = 6
# Create a new figure if no axis is provided.
if ax_provided is None:
fig, ax = plt.subplots(1,1)
ax.plot(years, melt, color="maroon", label = "Annual melt {0}".format("index" if melt_index_or_extent == "index" else "extent"))
melt_index_or_extent_lower = melt_index_or_extent.strip().lower()
if include_ylabel:
if melt_index_or_extent_lower == "index":
ax.set_ylabel("Melt Index (10$^{0}$ km$^2\cdot$days)".format(figure_exp))
# ax.set_ylabel("Melt Index (million km$^2$ days)")
elif melt_index_or_extent_lower == "extent":
ax.set_ylabel("Melt Extent (10$^{0}$ km$^2$)".format(figure_exp))
else:
raise ValueError("Unknown value for parameter 'melt_index_or_extent': {0}".format(melt_index_or_extent))
ax.tick_params(direction="in", bottom=True, left=True, right=True, top=False, labeltop=False, labelright=False, which="major")
ax.tick_params(direction="in", bottom=True, which="minor")
ax.tick_params(axis='x', length=4, which="major")
ax.tick_params(axis='x', length=2, which="minor")
if include_name_in_title:
region_name = antarctic_regions_dict[region_n]
ax.set_title(region_name)
# Limit lower-bounds to zero
ylim = ax.get_ylim()
ax.set_ylim(max(ylim[0], 0), ylim[1])
# Force the y-axis to only use integers (this tends to give us better scaling)
if ylim[1] > 8:
ax.yaxis.set_major_locator(matplotlib.ticker.MaxNLocator(integer=True))
# Turn on the minor ticks for the years.
ax.xaxis.set_minor_locator(matplotlib.ticker.MultipleLocator(base=1))
# ax.xaxis.grid(True, which='minor')
# Run a linear-fit OLS model on the data.
x = statsmodels.api.add_constant(years)
model = statsmodels.api.OLS(melt, x)
results = model.fit()
# If go into all this if we've indicated we might want to plot a trendline.
if include_trendline or include_trendline_only_if_significant:
# print(results.params)
# print(results.pvalues)
pval_int, pval_slope = results.pvalues
intercept, slope = results.params
# fit_func = numpy.poly1d((slope, intercept))
if print_trendline_summary:
print("\n")
print("============", antarctic_regions_dict[region_n] + ",", melt_index_or_extent, "==============")
print(results.summary())
if include_trendline or (pval_slope <= 0.05 and include_trendline_only_if_significant):
st, data, ss2 = summary_table(results, alpha=0.05)
fittedvalues = data[:, 2]
# predict_mean_se = data[:, 3]
predict_mean_ci_low, predict_mean_ci_upp = data[:, 4:6].T
predict_ci_low, predict_ci_upp = data[:, 6:8].T
# Put the p-value in the legend text.
p_value_text = ("{0:0.3f}" if (pval_slope > 0.001) else "{0:0.1e}").format(pval_slope)
# ax.plot(years, fit_func(years), color="blue", label = r"Linear Trend (\textit{p=" + p_value_text + "})")
# ax.plot(years, fit_func(years), color="blue", label = r"Linear Trend ($\it{p=" + p_value_text + "}$)")
ax.plot(years, fittedvalues, color="blue", label = r"Linear trend ($\it{p=" + p_value_text + "}$)")
if add_confidence_intervals:
# Regression errors, Y minus Y_fit
# y_err = melt - fit_func(years)
# Calculate confidence intervals
# p_x, confs = CI.conf_calc(years, y_err, c_limit=0.975, test_n=50)
# Calculate the lines for plotting:
# The fit line, and lower and upper confidence bounds
# p_y, lower, upper = CI.ylines_calc(p_x, confs, fit_func)
# plot confidence limits
# ax.plot(p_x, lower, 'c--',
ax.plot(years, predict_mean_ci_low, color='blue', linestyle='--',
label='95% confidence interval',
# label='95\% Confidence Interval',
alpha=0.5,
linewidth=0.8)
# ax.plot(p_x, upper, 'c--',
ax.plot(years, predict_mean_ci_upp, color='blue', linestyle='--',
label=None,
alpha=0.5,
linewidth=0.8)
if add_prediction_intervals:
ax.plot(years, predict_ci_low, color="red", linestyle='--',
label='95% prediction interval',
# label='95\% Confidence Interval',
alpha=0.5,
linewidth=0.5)
# ax.plot(p_x, upper, 'c--',
ax.plot(years, predict_ci_upp, color="red", linestyle='--',
label=None,
alpha=0.5,
linewidth=0.5)
# The prediction intervals are quite wide. Rescale the y-limits
# to be no more than 10% above/below the max/min of the data,
# even if it makes the prediction intervals trail off the figure
# a bit.
ylim = ax.get_ylim()
if (ylim[0] < 0) or (ylim[0] < (min(melt) - 0.1*(max(melt) - min(melt)))):
ax.set_ylim(max(0, min(melt)- 0.1*(max(melt) - min(melt))), ylim[1])
ylim = ax.get_ylim()
if (ylim[1] > (max(melt) + 0.1*(max(melt) - min(melt)))):
ax.set_ylim(ylim[0], (max(melt) + 0.1*(max(melt) - min(melt))))
if include_legend_if_significant:
ax.legend(fontsize="small", labelspacing=0.1, framealpha=0.95)
if ax_provided is None:
fig.tight_layout()
if fname_template is None:
plt.show()
else:
fname = fname_template.format(region_n)
fig.savefig(fname, dpi=dpi)
if verbose:
print(fname, "written.")
plt.close(fig)
return results
def Fig1():
OUT = open(mydir + 'output/PerDataset.txt','w+')
""" This code generates a 4 plot figure of diversity properties (rarity,
dominance, evenness, richness) versus total abundance, for each dataset.
This code also generates a .txt file of results for the regression
analyses. """
datasets = []
#GoodNames = ['TARA', 'HUMAN', 'BOVINE', 'LAUB', 'SED', 'CHU', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO', 'HMP', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA', 'EMPclosed', 'EMPopen']
GoodNames = ['HMP']
for name in os.listdir(mydir +'data/micro'):
if name in GoodNames: pass
else: continue
#path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir+'data/micro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'micro', num_lines])
print name, num_lines
for name in os.listdir(mydir +'data/macro'):
if name in GoodNames: pass
else: continue
#path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir+'data/macro/'+name+'/'+name+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'macro', num_lines])
print name, num_lines
metrics = ['Rarity, '+r'$log_{10}$',
'Dominance, '+r'$log_{10}$',
'Evenness, ' +r'$log_{10}$',
'Richness, ' +r'$log_{10}$']
#OUT = open(mydir + 'output/SummaryPerDataset_NoMicrobe1s.txt','w+')
OUT = open(mydir + 'output/SummaryPerDataset.txt','w+')
for dataset in datasets:
fig = plt.figure()
for index, i in enumerate(metrics):
metric = i
fig.add_subplot(2, 2, index+1)
fs = 10 # font size used across figures
IntList, CoefList, R2List, metlist = [[], [], [], []]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
its = 1
f = list()
for n in range(its):
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
radDATA = []
name, kind, numlines = dataset
lines = []
lines = np.random.choice(range(1, numlines+1), numlines, replace=False)
#if numlines > 1000:
# lines = np.random.choice(range(1, numlines+1), 1000, replace=True)
#else:
# lines = np.random.choice(range(1, numlines+1), numlines, replace=False)
#path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData_NoMicrobe1s.txt'
path = mydir+'data/'+kind+'/'+name+'/'+name+'-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
tN = 0
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
N = float(N)
tN += N
S = float(S)
#if S < 2 or N < 10: continue
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
ESimplist.append(float(np.log10(float(ESimp))))
KindList.append(kind)
BPlist.append(float(BP))
NmaxList.append(float(np.log10(float(Nmax))))
# log-modulo transformation of skewnness
lms = np.log10(np.abs(float(skew)) + 1)
if skew < 0: lms = lms * -1
rareSkews.append(float(lms))
print 'total number of reads in', name, ':',
print '%.3e' % tN
sys.exit()
if index == 0: metlist = list(rareSkews)
elif index == 1: metlist = list(NmaxList)
elif index == 2: metlist = list(ESimplist)
elif index == 3: metlist = list(Slist)
# Simple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
f = smf.ols('y ~ N', d).fit()
IntList.append(f.params[0])
CoefList.append(f.params[1])
R2List.append(f.rsquared)
PIx = list(Nlist)
st, data, ss2 = summary_table(f, alpha=0.05)
# ss2: Obs, Dep Var Population, Predicted Value, Std Error Mean Predict,
# Mean ci 95% low, Mean ci 95% upp, Predict ci 95% low, Predict ci 95% upp,
# Residual, Std Error Residual, Student Residual, Cook's D
Fitted = data[:,2]
predict_mean_se = data[:,3]
CiL, CiH = data[:,4:6].T
PiL, PiH = data[:,6:8].T
PIx, Fitted, CiH, CiL = zip(*sorted(zip(PIx, Fitted, CiH, CiL)))
plt.scatter(Nlist, metlist, color = 'SkyBlue', alpha= 1 , s = 4, linewidths=0.5, edgecolor='Steelblue')
plt.fill_between(PIx, CiL, CiH, color='b', lw=0.0, alpha=0.3)
Int = round(np.mean(IntList), 2)
Coef = round(np.mean(CoefList), 2)
R2 = round(np.mean(R2List), 3)
print dataset, metric, Int, Coef, R2
x = min(Nlist)
y = 1.1*max(metlist)
plt.scatter([0],[-1], color = 'SkyBlue', alpha = 1, s=10, linewidths=0.9, edgecolor='Steelblue', label= metric+' = '+str(round(10**Int, 2))+'*'+r'$N$'+'$^{'+str(round(Coef, 2))+'}$'+'\n'+r'$R^2$' + '=' +str(R2) +' (n='+str(len(PIx))+')')
if index == 2:
leg = plt.legend(loc=3,prop={'size':fs-1})
leg.draw_frame(False)
else:
leg = plt.legend(loc=2,prop={'size':fs-1})
leg.draw_frame(False)
plt.ylim(min(metlist), max(metlist)*1.1)
plt.xlim(min(Nlist), max(Nlist))
plt.xlabel('Total abundance, ' + r'$log_{10}(N)$', fontsize=fs)
plt.ylabel(metric, fontsize=fs)
plt.tick_params(axis='both', which='major', labelsize=fs-3)
metrix = ['rarity', 'dominance', 'evenness', 'richness']
print>>OUT, name, kind, metrix[index], np.mean(PIx), np.mean(Slist), Int, Coef
fig.suptitle(dataset[0], fontsize=fs+2)
#plt.subplots_adjust(wspace=0.4, hspace=0.4)
#plt.savefig(mydir+'/figs/appendix/Fig1/PerDataset/Locey_Lennon_2015_'+name+'_NoMicrobeSingletons.png', dpi=600, bbox_inches = "tight")
plt.savefig(mydir+'/figs/appendix/Fig1/PerDataset/Locey_Lennon_2015_'+name+'.png', dpi=600, bbox_inches = "tight")
#plt.show()
#plt.close()
OUT.close()
return
def Fig1(ref, Ones):
datasets = []
if ref == 'ClosedRef':
GoodNames = [
'MGRAST', 'HMP', 'EMPclosed', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'
]
if ref == 'OpenRef':
GoodNames = [
'MGRAST', 'HMP', 'EMPopen', 'BBS', 'CBC', 'MCDB', 'GENTRY', 'FIA'
]
for name in os.listdir(mydir + 'data/micro'):
if name in GoodNames: pass
else: continue
if Ones == 'N':
path = mydir + 'data/micro/' + name + '/' + name + '-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y':
path = mydir + 'data/micro/' + name + '/' + name + '-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'micro', num_lines])
print name, num_lines
for name in os.listdir(mydir + 'data/macro'):
if name in GoodNames: pass
else: continue
if Ones == 'N':
path = mydir + 'data/macro/' + name + '/' + name + '-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y':
path = mydir + 'data/macro/' + name + '/' + name + '-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, 'macro', num_lines])
print name, num_lines
metrics = ['log-modulo skewness', 'log-skew']
fig = plt.figure()
for index, i in enumerate(metrics):
metric = i
fig.add_subplot(2, 2, index + 1)
fs = 10 # font size used across figures
MicIntList, MicCoefList, MacIntList, MacCoefList, R2List, metlist = [
[], [], [], [], [], []
]
Nlist, Slist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList = [
[], [], [], [], [], [], [], [], []
]
EvarList, EQList, OList = [[], [], []]
SkewList, LogSkewList = [[], []]
its = 1000
for n in range(its):
#print n, metric
Nlist, Slist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList = [
[], [], [], [], [], [], [], [], []
]
EvarList, EQList, OList = [[], [], []]
SkewList, LogSkewList = [[], []]
numMac = 0
numMic = 0
radDATA = []
for dataset in datasets:
name, kind, numlines = dataset
lines = []
if name == 'EMPclosed' or name == 'EMPopen':
lines = np.random.choice(range(1, numlines + 1),
100,
replace=True)
elif kind == 'micro':
lines = np.random.choice(range(1, numlines + 1),
100,
replace=True)
else:
lines = np.random.choice(range(1, numlines + 1),
60,
replace=True)
if Ones == 'N':
path = mydir + 'data/' + kind + '/' + name + '/' + name + '-SADMetricData_NoMicrobe1s.txt'
elif Ones == 'Y':
path = mydir + 'data/' + kind + '/' + name + '/' + name + '-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
KindList.append(kind)
N = float(N)
S = float(S)
if S < 10 or N < 10: continue # Min species richness
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
# Rarity
lms = np.log10(np.abs(float(skew)) + 1)
if skew < 0: lms = lms * -1
SkewList.append(float(lms))
LogSkewList.append(float(logskew))
if kind == 'micro':
numMic += 1
klist.append('b')
if kind == 'macro':
klist.append('r')
numMac += 1
if index == 0: metlist = list(SkewList)
elif index == 1: metlist = list(LogSkewList)
# Multiple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
f = smf.ols('y ~ N * Kind', d).fit()
MacIntList.append(f.params[0])
MacCoefList.append(f.params[2])
if f.pvalues[1] < 0.05:
MicIntList.append(f.params[1] + f.params[0])
else:
MicIntList.append(f.params[0])
if f.pvalues[3] < 0.05:
MicCoefList.append(f.params[3] + f.params[2])
else:
MicCoefList.append(f.params[2])
R2List.append(f.rsquared)
MacListX = []
MacListY = []
MicListX = []
MicListY = []
for j, k in enumerate(KindList):
if k == 'micro':
MicListX.append(Nlist[j])
MicListY.append(metlist[j])
elif k == 'macro':
MacListX.append(Nlist[j])
MacListY.append(metlist[j])
MacPIx, MacFitted, MicPIx, MicFitted = [[], [], [], []]
macCiH, macCiL, micCiH, micCiL = [[], [], [], []]
lm = smf.ols('y ~ N * Kind', d).fit()
#print metric, '\n', lm.summary()
#f1 = smf.ols('y ~ N', d).fit()
#print metric, '\n', f1.summary()
st, data, ss2 = summary_table(lm, alpha=0.05)
fittedvalues = data[:, 2]
predict_mean_se = data[:, 3]
predict_mean_ci_low, predict_mean_ci_upp = data[:, 4:6].T
predict_ci_low, predict_ci_upp = data[:, 6:8].T
for j, kval in enumerate(KindList):
if kval == 'macro':
macCiH.append(predict_mean_ci_upp[j])
macCiL.append(predict_mean_ci_low[j])
MacPIx.append(Nlist[j])
MacFitted.append(f.fittedvalues[j])
elif kval == 'micro':
micCiH.append(predict_mean_ci_upp[j])
micCiL.append(predict_mean_ci_low[j])
MicPIx.append(Nlist[j])
MicFitted.append(f.fittedvalues[j])
MicPIx, MicFitted, micCiH, micCiL = zip(
*sorted(zip(MicPIx, MicFitted, micCiH, micCiL)))
MacPIx, MacFitted, macCiH, macCiL = zip(
*sorted(zip(MacPIx, MacFitted, macCiH, macCiL)))
for i in range(len(MicListX)):
plt.scatter(MacListX[i],
MacListY[i],
color='LightCoral',
alpha=1,
s=4,
linewidths=0.5,
edgecolor='Crimson')
plt.scatter(MicListX[i],
MicListY[i],
color='SkyBlue',
alpha=1,
s=4,
linewidths=0.5,
edgecolor='Steelblue')
plt.fill_between(MacPIx, macCiL, macCiH, color='r', lw=0.0, alpha=0.3)
plt.fill_between(MicPIx, micCiL, micCiH, color='b', lw=0.0, alpha=0.3)
MicInt = round(np.mean(MicIntList), 2)
MicCoef = round(np.mean(MicCoefList), 2)
MacInt = round(np.mean(MacIntList), 2)
MacCoef = round(np.mean(MacCoefList), 2)
r2 = round(np.mean(R2List), 2)
if index == 0:
plt.ylim(0, 2.5)
plt.xlim(0, 7)
plt.text(0.3,
2.2,
r'$micro$' + ' = ' + str(round(MicInt, 2)) + '*' +
r'$N$' + '$^{' + str(round(MicCoef, 2)) + '}$',
fontsize=fs - 1,
color='Steelblue')
plt.text(0.3,
2.0,
r'$macro$' + ' = ' + str(round(MacInt, 2)) + '*' +
r'$N$' + '$^{' + str(round(MacCoef, 2)) + '}$',
fontsize=fs - 1,
color='Crimson')
plt.text(0.3,
1.7,
r'$R^2$' + '=' + str(round(r2, 3)),
fontsize=fs - 1,
color='k')
if index == 1:
plt.ylim(-1, 4.5)
plt.xlim(0, 7)
plt.text(0.3,
4.0,
r'$micro$' + ' = ' + str(round(MicInt, 2)) + '*' +
r'$N$' + '$^{' + str(round(MicCoef, 2)) + '}$',
fontsize=fs - 1,
color='Steelblue')
plt.text(0.3,
3.5,
r'$macro$' + ' = ' + str(round(MacInt, 2)) + '*' +
r'$N$' + '$^{' + str(round(MacCoef, 2)) + '}$',
fontsize=fs - 1,
color='Crimson')
plt.text(0.3,
2.9,
r'$R^2$' + '=' + str(round(r2, 3)),
fontsize=fs - 1,
color='k')
plt.xlabel('Number of reads or individuals, ' + '$log$' + r'$_{10}$',
fontsize=fs)
plt.ylabel(metric, fontsize=fs)
plt.tick_params(axis='both', which='major', labelsize=fs - 1)
plt.subplots_adjust(wspace=0.4, hspace=0.4)
if ref == 'OpenRef' and Ones == 'N':
plt.savefig(
mydir +
'/figs/appendix/Rarity/SupplementaryRarityFig-OpenRef_NoMicrobe1s.png',
dpi=600,
bbox_inches="tight")
elif ref == 'OpenRef' and Ones == 'Y':
plt.savefig(mydir +
'/figs/appendix/Rarity/SupplementaryRarityFig-OpenRef.png',
dpi=600,
bbox_inches="tight")
elif ref == 'ClosedRef' and Ones == 'Y':
plt.savefig(
mydir +
'/figs/appendix/Rarity/SupplementaryRarityFig-ClosedRef.png',
dpi=600,
bbox_inches="tight")
elif ref == 'ClosedRef' and Ones == 'N':
plt.savefig(
mydir +
'/figs/appendix/Rarity/SupplementaryRarityFig-ClosedRef_NoMicrobe1s.png',
dpi=600,
bbox_inches="tight")
#plt.show()
return
def scatter_interaction(ax, x, y, groups, colors, ms=5, labels=None, title=None, legend=False, formula='y ~ x', legend_loc='best'):
# Here are the imports
import numpy as np
import matplotlib.pylab as plt
import pandas as pd
from scipy.stats import pearsonr
from statsmodels.sandbox.regression.predstd import wls_prediction_std
from statsmodels.formula.api import ols
from statsmodels.stats.outliers_influence import summary_table
# If you haven't already been given an axis on which to plot, then
# create a new figure
if not ax:
fig = plt.figure(figsize = (5,4))
ax = fig.add_subplot(111)
marker_styles = [ 'o', '^', 'D', 's', '*' ] * len(groups)
line_styles = ['--', '-', '-.', ':'] * len(groups)
# Loop through all the groups
for i, x_i, y_i, c_i, g_i, m_i, l_i in zip(range(len(groups)), x, y, colors, groups, marker_styles, line_styles):
# Scatter each with the appropriate colors
ax.scatter(x_i, y_i, c=c_i, edgecolor=c_i, alpha=0.8, s=ms, marker=m_i, zorder=7*i)
# Now calculate the linear correlation between x and y
# for each group
# Heavily stolen from:
# http://www.students.ncl.ac.uk/tom.holderness/software/pythonlinearfit
#z = np.polyfit(x_i,y_i,1)
#p = np.poly1d(z)
#fit = p(x_i)
#c_x = [np.min(x_i),np.max(x_i)]
#c_y = [p(np.min(x_i)), p(np.max(x_i))]
df2 = pd.DataFrame({ 'x' : x_i, 'y' : y_i })
df2.sort('x', inplace=True)
lm = ols(formula, df2).fit()
ps = [ '{:2.4f}'.format(p) for p in lm.pvalues[1:] ]
print ' {}, r2 = {}, p(s) = {}'.format(g_i, lm.rsquared, ', '.join(ps))
prstd, iv_l, iv_u = wls_prediction_std(lm)
iv_l = np.array(iv_l)
iv_u = np.array(iv_u)
fit_y = np.array(lm.fittedvalues)
st, data, ss2 = summary_table(lm, alpha=0.05)
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
#pl.plot(x, y, 'k-')
#pl.plot(x, fit_y, 'r--')
#pl.fill_between(x, iv_l, iv_u, alpha=0.2)
# Get the r and p values
#r, p = pearsonr(x_i, y_i)
#label = '{} r: {: .2g} p: {: .2g}'.format(g_i, r, p)
# Now plot
ax.plot(df2.x.values, fit_y, c=c_i, linestyle = '-', linewidth = ms/25.0, zorder=6*i, label=g_i)
ax.plot(df2.x.values, predict_mean_ci_low, c=c_i, linestyle = '-', linewidth = ms/50.0, zorder=3*i)
ax.plot(df2.x.values, predict_mean_ci_upp, c=c_i, linestyle = '-', linewidth = ms/50.0, zorder=2*i)
ax.fill_between(df2.x.values, predict_mean_ci_upp, predict_mean_ci_low, alpha=0.3, facecolor=c_i, interpolate=True, zorder=1*i)
if legend:
# Add the legend
leg = ax.legend(loc=legend_loc, fancybox=True, fontsize=ms/2.)
leg.get_frame().set_alpha(0)
# Set the y limits
# This is to deal with very small numbers (the MaxNLocator gets all turned around!)
# Concatenate all the y data:
y_all = y[0]
if len(y) > 1:
for k in range(1,len(y)):
y_all = np.concatenate([y_all, y[k]])
max_y = np.max(y_all)
min_y = np.min(y_all)
buffer = ( max_y - min_y ) / 10
upper = max_y + buffer
lower = min_y - buffer
ax.set_ybound(upper, lower)
# Set the axis labels
ax.set_ylabel(labels[1], fontsize=ms/2.0)
ax.set_xlabel(labels[0], fontsize=ms/2.0)
for item in (ax.get_xticklabels() + ax.get_yticklabels()):
item.set_fontsize(ms/2.0)
# Adjust the power limits so that you use scientific notation on the y axis
plt.ticklabel_format(style='sci', axis='y')
ax.yaxis.major.formatter.set_powerlimits((-3,3))
plt.rc('font', **{'size':ms/2.0})
if title:
# Set the overall title
ax.set_title(title)
plt.tight_layout()
return ax
def Fig2(condition, ones, sampling):
""" A figure demonstrating a strong abundance relationship across 30
orders of magnitude in total abundance. The abundance of the most abundant
species scales in a log-log fashion with the total abundance of the sample
or system. """
tail = str()
if ones is False:
tail = '-SADMetricData_NoMicrobe1s.txt'
elif ones is True:
tail = '-SADMetricData.txt'
datasets = []
GoodNames = []
emp = str()
if condition == 'open': emp = 'EMPopen'
elif condition == 'closed': emp = 'EMPclosed'
GoodNames = [emp, 'TARA', 'HMP', 'BIGN', 'BOVINE', 'CHU', 'LAUB', 'SED', 'HUMAN', 'CHINA', 'CATLIN', 'FUNGI']
fs = 13 # font size used across figures
Nlist, NmaxList, klist, datasets, radDATA = [[],[],[],[],[]]
for name in os.listdir(mydir +'data/micro'):
#if name in BadNames: continue
if name in GoodNames: pass
else: continue
path = mydir+'data/micro/'+name+'/'+name+tail
numlines = sum(1 for line in open(path))
print name, numlines
datasets.append([name, 'micro', numlines])
for dataset in datasets:
name, kind, numlines = dataset
lines = []
small = ['BIGN', 'BOVINE', 'CHU', 'LAUB', 'SED']
big = ['HUMAN', 'CHINA', 'CATLIN', 'FUNGI', 'HYDRO']
if name in small:
lines = np.random.choice(range(1, numlines+1), 1000, replace=True)
elif name in big:
lines = np.random.choice(range(1, numlines+1), 2500, replace=True)
elif name == 'TARA':
lines = np.random.choice(range(1, numlines+1), 2500, replace=True)
else:
lines = np.random.choice(range(1, numlines+1), 2500, replace=True)
path = mydir+'data/micro/'+name+'/'+name+tail
for line in lines:
data = linecache.getline(path, line)
radDATA.append(data)
klist.append('DarkCyan')
for data in radDATA:
data = data.split()
name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
N = float(N)
S = float(S)
#if S < 10 or N < 11: continue # Min species richness
Nlist.append(float(np.log10(float(N))))
NmaxList.append(float(np.log10(float(Nmax))))
klist.append('DarkCyan')
metric = 'Dominance, '+'$log$'+r'$_{10}$'
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
Nlist, NmaxList = zip(*sorted(zip(Nlist, NmaxList)))
Nlist = list(Nlist)
NmaxList = list(NmaxList)
# Regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(NmaxList)
f = smf.ols('y ~ N', d).fit()
R2 = f.rsquared
pval = f.pvalues
intercept = f.params[0]
slope = f.params[1]
#print f.summary()
#print intercept, slope
X = np.linspace(6, 40, 100)
Y = f.predict(exog=dict(N=X))
Nlist2 = Nlist + X.tolist()
NmaxList2 = NmaxList + Y.tolist()
d = pd.DataFrame({'N': list(Nlist2)})
d['y'] = list(NmaxList2)
f = smf.ols('y ~ N', d).fit()
st, data, ss2 = summary_table(f, alpha=0.05)
fittedvalues = data[:,2]
pred_mean_se = data[:,3]
pred_mean_ci_low, pred_mean_ci_upp = data[:,4:6].T
pred_ci_low, pred_ci_upp = data[:,6:8].T
label1 = 'Dominance scaling law for microbial data compilation'
label2 = 'Ranges of published $N_{max}$ and $N$'
plt.fill_between(Nlist2, pred_ci_low, pred_ci_upp, color='r', lw=0.5, alpha=0.2)
plt.text(2, 22, r'$N_{max}$'+ ' = '+str(round(10**intercept,2))+'*'+r'$N$'+'$^{'+str(round(slope,2))+'}$', fontsize=fs+4, color='Crimson', alpha=0.9)
plt.text(2, 19, r'$r^2$' + ' = ' +str("%.2f" % R2), fontsize=fs+4, color='0.2')
plt.plot(X.tolist(), Y.tolist(), '--', c='red', lw=2, alpha=0.8, color='Crimson', label=label1)
print 'r-squared and slope for RADs w/out inferred:', round(R2, 3), round(slope,3)
plt.hexbin(Nlist, NmaxList, mincnt=1, gridsize = 50, bins='log', cmap=plt.cm.Reds_r) #
#plt.scatter(Nlist, NmaxList, color = 'LightCoral', alpha= 0.6 , s = 10, linewidths=0.5, edgecolor='Crimson')
GO = np.log10([360.0*(10**26), 1010.0*(10**26)]) # estimated open ocean bacteria; Whitman et al. 1998
Pm = np.log10([2.8*(10**27), 3.0*(10**27)]) # estimated Prochlorococcus; Flombaum et al. 2013
Syn = np.log10([6.7*(10**26), 7.3*(10**26)]) # estimated Synechococcus; Flombaum et al. 2013
Earth = np.log10([9.2*(10**29), 31.7*(10**29)]) # estimated bacteria on Earth; Kallmeyer et al. 2012
SAR11 = np.log10([2.0*(10**28), 2.0*(10**28)]) # estimated percent abundance of SAR11; Morris et al. (2002)
HGx = np.log10([0.5*(10**14), 1.5*(10**14)]) # estimated bacteria in Human gut; Berg (1996)
HGy = np.log10([0.05*(10**min(HGx)), 0.15*(10**max(HGx))]) # estimated most abundant bacteria in Human gut; Turnbaugh et al. (2009), & Dethlefsen et al. (2008)
COWx = np.log10([0.5*2.226*(10**15), 1.5*2.226*(10**15)]) # estimated bacteria in Cow rumen; LOW: HIGH: Whitman et al. (1998)
COWy = np.log10([0.09*(10**min(COWx)), .15*(10**max(COWx))]) # estimated dominance in Cow rumen; Stevenson and Weimer (2006)
c = '0.2'
## EARTH
x = [np.mean(Earth)]
x_range = (max(Earth) - min(Earth))/2.0
y = [np.mean([min(Pm), max(SAR11)])]
y_range = (max(SAR11) - min(Pm))/2.0
ax.text(8.5, max(SAR11)+0.2, r'$Prochlorococcus$ and Pelagibacterales', fontsize=fs+2, color = 'k')
ax.text(max(Earth)+0.5, 26, 'Earth microbiome', fontsize=fs+2, color = 'k', rotation = 90)
ax.axhline(y, 0, 0.90, ls = '--', c = '0.4')
ax.axvline(x, 0, 0.85, ls = '--', c = '0.4')
plt.errorbar(x, y, xerr=x_range, yerr=y_range, color='k', linewidth=1, label=label2)
c = '0.4'
## GLOBAL OCEAN
x = [np.mean(GO)]
x_range = (max(GO) - min(GO))/2.0
y = [np.mean(Pm)]
y_range = (max(SAR11) - min(Pm))/2.0
ax.text(7.5, min(Pm)-1.35, r'$Synechococcus$ and $Prochlorococcus$', fontsize=fs+2, color = 'k')
ax.text(min(GO)-1, 22, 'Non-sediment ocean bacteria', fontsize=fs+2, color = 'k', rotation = 90)
ax.axhline(y, 0, 0.85, ls = '--', c = '0.4')
ax.axvline(x, 0, 0.83, ls = '--', c = '0.4')
plt.errorbar(x, y, xerr=x_range, yerr=y_range, color='k', linewidth=1)
## HUMAN GUT
x = [np.mean(HGx)]
x_range = (max(HGx) - min(HGx))/2.0
y = [np.mean(HGy)]
y_range = (max(HGy) - min(HGy))/2.0
ax.text(4, min(HGy)-1, 'Human gut', fontsize=fs+2, color = 'k')
ax.text(min(HGx)-1, 8, 'Human gut', fontsize=fs+2, color = 'k', rotation = 90)
ax.axhline(y, 0, 0.40, ls = '--', c = '0.4')
ax.axvline(x, 0, 0.38, ls = '--', c = '0.4')
plt.errorbar(x, y, xerr=x_range, yerr=y_range, color='k', linewidth=1)
## COW RUMEN
x = [np.mean(COWx)]
x_range = (max(COWx) - min(COWx))/2.0
y = [np.mean(COWy)]
y_range = (max(COWy) - min(COWy))/2.0
ax.text(7, max(COWy)+0.3, '$Prevotella$', fontsize=fs+2, color = 'k')
ax.text(max(COWx)+0.4, 11.2, 'Cow rumen', fontsize=fs+2, color = 'k', rotation = 90)
ax.axhline(y, 0, 0.41, ls = '--', c = '0.4')
ax.axvline(x, 0, 0.43, ls = '--', c = '0.4')
plt.errorbar(x, y, xerr=x_range, yerr=y_range, color='k', linewidth=1)
ax.text(5, -4.2, 'Number of reads or total abundance, '+ '$log$'+r'$_{10}$', fontsize=fs+4)
ax.text(-2.5, 22, metric, fontsize=fs+4, rotation=90)
plt.plot([0,32],[0,32], ls = '--', lw=2, c='0.7')
#ax.text(18, 21, '1:1 line', fontsize=fs*1.0, rotation=40, color='0.7')
plt.xlim(1, 33)
plt.ylim(0, 32)
plt.legend(bbox_to_anchor=(-0.015, 1, 1.025, .2), loc=10, ncol=1,
mode="expand",prop={'size':fs+1}, numpoints=1)
if ones == False:
plt.savefig(mydir+'/figs/Fig2/Locey_Lennon_2015_Fig2-'+condition+'_NoSingletons_'+str(sampling)+'.pdf', dpi=300, bbox_inches = "tight")
if ones == True:
plt.savefig(mydir+'/figs/Fig2/Locey_Lennon_2015_Fig2-'+condition+'_'+str(sampling)+'.pdf', dpi=300, bbox_inches = "tight")
#plt.savefig(mydir+'/figs/Fig2/figure2-v2.pdf', dpi=300, bbox_inches = "tight")
#plt.show()
return
leg.get_frame().set_alpha(0.5) #, fontsize='small')
ltext = leg.get_texts() # all the text.Text instance in the legend
plt.setp(ltext, fontsize='small') # the legend text fontsize
print oi.reset_ramsey(res, degree=3)
#note, constant in last column
for i in range(1):
print oi.variance_inflation_factor(res.model.exog, i)
infl = oi.OLSInfluence(res_ols)
print infl.resid_studentized_external
print infl.resid_studentized_internal
print infl.summary_table()
print oi.summary_table(res, alpha=0.05)[0]
'''
>>> res.resid
array([ 4.28571429, 4. , 0.57142857, -3.64285714,
-4.71428571, 1.92857143, 10. , -6.35714286,
-11. , -1.42857143, 1.71428571, 4.64285714])
>>> infl.hat_matrix_diag
array([ 0.10084034, 0.11764706, 0.28571429, 0.20168067, 0.10084034,
0.16806723, 0.11764706, 0.08403361, 0.11764706, 0.28571429,
0.33613445, 0.08403361])
>>> infl.resid_press
array([ 4.76635514, 4.53333333, 0.8 , -4.56315789,
-5.24299065, 2.31818182, 11.33333333, -6.94036697,
-12.46666667, -2. , 2.58227848, 5.06880734])
>>> infl.ess_press
def plot_OLS(ax, target, Y, mode='unicolor'):
X = target
X = sm.add_constant(X)
model = sm.OLS(Y, X)
results = model.fit()
st, data, ss2 = summary_table(results, alpha=0.05)
fittedvalues = data[:, 2]
predict_mean_se = data[:, 3]
predict_mean_ci_low, predict_mean_ci_upp = data[:, 4:6].T
predict_ci_low, predict_ci_upp = data[:, 6:8].T
if mode == 'unicolor':
ax.scatter(target, Y, c='silver', linewidths=0, s=4)
else:
xy = np.row_stack([target, Y])
z = gaussian_kde(xy)(xy)
idx = z.argsort()
x, y, z = xy[0][idx], xy[1][idx], z[idx]
ax.scatter(x, y, c=z, s=4, cmap=pl.cm.inferno_r)
ax.plot(target, fittedvalues, 'r-', label='Least Square Regression', lw=2)
idx = np.argsort(predict_ci_low)
ax.plot(target[idx],
predict_ci_low[idx],
'r--',
lw=2,
label='95% confidence interval')
idx = np.argsort(predict_ci_upp)
ax.plot(target[idx], predict_ci_upp[idx], 'r--', lw=2)
mx = np.ceil(max(target.max(), fittedvalues.max()))
ax.plot([0, mx], [0, mx], 'k-')
ax.set_xlim(0, mx)
ax.set_ylim(0, mx)
ax.set_aspect(1)
ax.legend(loc='upper left')
ax.set_xlabel('AGB from map [Mg ha$^{-1}$]')
ax.set_ylabel('Reconstructed AGB [Mg ha$^{-1}$]')
nse = 1 - ((Y - target)**2).sum() / ((target - target.mean())**2).sum()
rmse = np.sqrt(((Y - target)**2).mean())
ax.text(
0.98,
0.02,
'y = %4.2fx + %4.2f\nR$^2$ = %4.2f; p < 0.001\nrmse = %4.1f Mg ha$^{-1}$ ; NSE = %4.2f'
% (results.params[1], results.params[0], results.rsquared, rmse, nse),
va='bottom',
ha='right',
transform=ax.transAxes)
idx = np.argsort(predict_ci_upp)
ax.plot(target[idx], predict_ci_upp[idx], 'r--', lw=2)
mx = np.ceil(max(target.max(), fittedvalues.max()))
ax.plot([0, mx], [0, mx], 'k-')
ax.set_xlim(0, mx)
ax.set_ylim(0, mx)
ax.set_aspect(1)
ax.legend(loc='upper left')
ax.set_xlabel('AGB from map [Mg ha$^{-1}$]')
ax.set_ylabel('Reconstructed AGB [Mg ha$^{-1}$]')
nse = 1 - ((Y - target)**2).sum() / ((target - target.mean())**2).sum()
rmse = np.sqrt(((Y - target)**2).mean())
ax.text(
0.98,
0.02,
'y = %4.2fx + %4.2f\nR$^2$ = %4.2f; p < 0.001\nrmse = %4.1f Mg ha$^{-1}$ ; NSE = %4.2f'
% (results.params[1], results.params[0], results.rsquared, rmse, nse),
va='bottom',
ha='right',
transform=ax.transAxes)
def Fig1(cutoffs, Ones):
datasets = []
GoodNames = ['LAUB', 'CHU', 'HYDRO', 'CATLIN']
for cutoff in cutoffs:
for name in GoodNames:
if Ones == 'N':
path = mydir+'data/micro/'+name+'/'+name+cutoff+'/'+name+cutoff+'-SADMetricData_NoMicrobe1s.txt'
if Ones == 'Y':
path = mydir+'data/micro/'+name+'/'+name+cutoff+'/'+name+cutoff+'-SADMetricData.txt'
num_lines = sum(1 for line in open(path))
datasets.append([name, cutoff, 'micro', num_lines])
print name, num_lines
metrics = ['Rarity, '+r'$log_{10}$',
'Dominance, '+r'$log_{10}$',
'Evenness, ' +r'$log_{10}$',
'Richness, ' +r'$log_{10}$']
fig = plt.figure()
for index, i in enumerate(metrics):
metric = i
fig.add_subplot(2, 2, index+1)
fs = 10 # font size used across figures
c97IntList, c97CoefList, c99IntList, c99CoefList, c95CoefList, c95IntList, R2List, metlist = [[], [], [], [], [], [], [], []]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
its = 100
for n in range(its):
#name, kind, N, S, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = [[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]]
Nlist, Slist, Evarlist, ESimplist, klist, radDATA, BPlist, NmaxList, rareSkews, KindList, StdList = [[], [], [], [], [], [], [], [], [], [], []]
radDATA = []
for dataset in datasets:
name, cutoff, kind, numlines = dataset
lines = []
lines = np.random.choice(range(1, numlines+1), numlines, replace=False)
if Ones == 'N':
path = mydir+'data/'+kind+'/'+name+'/'+name+cutoff+'/'+name+cutoff+'-SADMetricData_NoMicrobe1s.txt'
if Ones == 'Y':
path = mydir+'data/'+kind+'/'+name+'/'+name+cutoff+'/'+name+cutoff+'-SADMetricData.txt'
for line in lines:
data = linecache.getline(path, line)
dlist = cutoff+' '+data
radDATA.append(dlist)
for data in radDATA:
data = data.split()
cutoff, name, kind, N, S, Var, Evar, ESimp, EQ, O, ENee, EPielou, EHeip, BP, SimpDom, Nmax, McN, skew, logskew, chao1, ace, jknife1, jknife2, margalef, menhinick, preston_a, preston_S = data
N = float(N)
S = float(S)
#if S < 10 or N < 11: continue
Nlist.append(float(np.log10(N)))
Slist.append(float(np.log10(S)))
ESimplist.append(float(np.log10(float(ESimp))))
KindList.append(cutoff)
BPlist.append(float(BP))
NmaxList.append(float(np.log10(float(Nmax))))
# log-modulo transformation of skewnness
lms = np.log10(np.abs(float(skew)) + 1)
if skew < 0: lms = lms * -1
rareSkews.append(float(lms))
if cutoff == '99':
klist.append('Steelblue')
elif cutoff == '97':
klist.append('Crimson')
elif cutoff == '95':
klist.append('0.4')
if index == 0: metlist = list(rareSkews)
elif index == 1: metlist = list(NmaxList)
elif index == 2: metlist = list(ESimplist)
elif index == 3: metlist = list(Slist)
# Multiple regression
d = pd.DataFrame({'N': list(Nlist)})
d['y'] = list(metlist)
d['Kind'] = list(KindList)
f = smf.ols('y ~ N * Kind', d).fit()
print f.summary()
#print f.params
c95IntList.append(f.params[0])
c95CoefList.append(f.params[3])
if f.pvalues[1] < 0.05: c97IntList.append(f.params[1] + f.params[0])
else: c97IntList.append(f.params[0])
if f.pvalues[4] < 0.05: c97CoefList.append(f.params[4] + f.params[3])
else: c97CoefList.append(f.params[3])
if f.pvalues[2] < 0.05: c99IntList.append(f.params[2] + f.params[0])
else: c99IntList.append(f.params[0])
if f.pvalues[5] < 0.05: c99CoefList.append(f.params[5] + f.params[3])
else: c99CoefList.append(f.params[3])
R2List.append(f.rsquared)
c95PIx, c95Fitted = [[],[]]
c95CiH, c95CiL = [[],[]]
c97PIx, c97Fitted = [[],[]]
c97CiH, c97CiL = [[],[]]
c99PIx, c99Fitted = [[],[]]
c99CiH, c99CiL = [[],[]]
c95ListX = []
c95ListY = []
c97ListX = []
c97ListY = []
c99ListX = []
c99ListY = []
for j, k in enumerate(KindList):
if k == '99':
c99ListX.append(Nlist[j])
c99ListY.append(metlist[j])
if k == '97':
c97ListX.append(Nlist[j])
c97ListY.append(metlist[j])
if k == '95':
c95ListX.append(Nlist[j])
c95ListY.append(metlist[j])
print metric
lm = smf.ols('y ~ N * Kind', d).fit()
print lm.summary()
print '\n\n'
st, data, ss2 = summary_table(lm, alpha=0.05)
# ss2: Obs, Dep Var Population, Predicted Value, Std Error Mean Predict,
# Mean ci 95% low, Mean ci 95% upp, Predict ci 95% low, Predict ci 95% upp,
# Residual, Std Error Residual, Student Residual, Cook's D
#fittedvalues = data[:,2]
#predict_mean_se = data[:,3]
predict_mean_ci_low, predict_mean_ci_upp = data[:,4:6].T
predict_ci_low, predict_ci_upp = data[:,6:8].T
for j, kval in enumerate(KindList):
if kval == '99':
c99CiH.append(predict_mean_ci_upp[j])
c99CiL.append(predict_mean_ci_low[j])
c99PIx.append(Nlist[j])
c99Fitted.append(f.fittedvalues[j])
if kval == '97':
c97CiH.append(predict_mean_ci_upp[j])
c97CiL.append(predict_mean_ci_low[j])
c97PIx.append(Nlist[j])
c97Fitted.append(f.fittedvalues[j])
if kval == '95':
c95CiH.append(predict_mean_ci_upp[j])
c95CiL.append(predict_mean_ci_low[j])
c95PIx.append(Nlist[j])
c95Fitted.append(f.fittedvalues[j])
c99PIx, c99Fitted, c99CiH, c99CiL, c97PIx, c97Fitted, c97CiH, c97CiL, c95PIx, c95Fitted, c95CiH, c95CiL = zip(*sorted(zip(c99PIx, c99Fitted, c99CiH, c99CiL, c97PIx, c97Fitted, c97CiH, c97CiL, c95PIx, c95Fitted, c95CiH, c95CiL)))
plt.scatter(c99ListX, c99ListY, facecolor = 'none', alpha= 1 , s = 5, linewidths=0.5, edgecolor='Steelblue')
plt.scatter(c97ListX, c97ListY, facecolor = 'none', alpha= 1 , s = 5, linewidths=0.5, edgecolor='Crimson')
plt.scatter(c95ListX, c95ListY, facecolor = 'none', alpha= 1 , s = 5, linewidths=0.5, edgecolor='0.4')
#plt.fill_between(c99PIx, c99CiL, c99CiH, color='b', lw=0.0, alpha=0.3)
#plt.fill_between(c97PIx, c97CiL, c97CiH, color='r', lw=0.0, alpha=0.3)
#plt.fill_between(c95PIx, c95CiL, c95CiH, color='k', lw=0.0, alpha=0.3)
plt.plot(c99PIx, c99Fitted, color='b', ls='--', lw=1, alpha=0.8)
plt.plot(c97PIx, c97Fitted, color='r', ls='--', lw=1, alpha=0.8)
plt.plot(c95PIx, c95Fitted, color='0.2', ls='--', lw=1, alpha=0.8)
c99Int = round(np.mean(c99IntList), 2)
c99Coef = round(np.mean(c99CoefList), 2)
c97Int = round(np.mean(c97IntList), 2)
c97Coef = round(np.mean(c97CoefList), 2)
c95Int = round(np.mean(c95IntList), 2)
c95Coef = round(np.mean(c95CoefList), 2)
R2 = round(np.mean(R2List), 2)
if index == 0:
plt.ylim(-0.1, 2.0)
plt.xlim(1, 6)
plt.text(1.35, 1.7, r'$99%$'+ ' = '+str(round(10**c99Int,2))+'*'+r'$N$'+'$^{'+str(round(c99Coef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(1.35, 1.5, r'$97%$'+ ' = '+str(round(10**c97Int,2))+'*'+r'$N$'+'$^{'+str(round(c97Coef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(1.35, 1.3, r'$95%$'+ ' = '+str(round(10**c95Int,2))+'*'+r'$N$'+'$^{'+str(round(c95Coef,2))+'}$', fontsize=fs, color='0.3')
plt.text(1.35, 1.1, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
plt.scatter([0],[-1], color = 'none', alpha = 1, s=10, linewidths=0.9, edgecolor='Steelblue', label= '99% (n='+str(len(c99ListY))+')')
plt.scatter([0],[-1], color = 'none', alpha = 1, s=10, linewidths=0.9, edgecolor='Crimson', label= '97% (n='+str(len(c97ListY))+')')
plt.scatter([0],[-1], color = 'none', alpha = 1, s=10, linewidths=0.9, edgecolor='0.3', label= '95% (n='+str(len(c95ListY))+')')
plt.legend(bbox_to_anchor=(-0.04, 1.05, 2.48, .2), loc=10, ncol=3, mode="expand",prop={'size':fs+2})
elif index == 1:
plt.plot([0,7],[0,7], ls = '--', lw=1, c='0.7')
#ax.text(18, 21, '1:1 line', fontsize=fs*1.0, rotation=40, color='0.7')
plt.ylim(0, 6)
plt.xlim(1, 6)
plt.text(1.35, 5.1, r'$99%$'+ ' = '+str(round(10**c99Int,2))+'*'+r'$N$'+'$^{'+str(round(c99Coef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(1.35, 4.6, r'$97%$'+ ' = '+str(round(10**c97Int,2))+'*'+r'$N$'+'$^{'+str(round(c97Coef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(1.35, 4.1, r'$95%$'+ ' = '+str(round(10**c95Int,2))+'*'+r'$N$'+'$^{'+str(round(c95Coef,2))+'}$', fontsize=fs, color='0.3')
plt.text(1.35, 3.6, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
elif index == 2:
plt.ylim(-3.0, 0.0)
plt.xlim(1, 6)
plt.text(1.35, -1.8, r'$99%$'+ ' = '+str(round(10**c99Int,2))+'*'+r'$N$'+'$^{'+str(round(c99Coef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(1.35, -2.1, r'$97%$'+ ' = '+str(round(10**c97Int,2))+'*'+r'$N$'+'$^{'+str(round(c97Coef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(1.35, -2.4, r'$95%$'+ ' = '+str(round(10**c95Int,2))+'*'+r'$N$'+'$^{'+str(round(c95Coef,2))+'}$', fontsize=fs, color='0.3')
plt.text(1.35, -2.7, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
elif index == 3:
plt.ylim(0.9, 4.5)
plt.xlim(1, 6)
plt.text(1.35, 3.9, r'$99%$'+ ' = '+str(round(10**c99Int,2))+'*'+r'$N$'+'$^{'+str(round(c99Coef,2))+'}$', fontsize=fs, color='Steelblue')
plt.text(1.35, 3.5, r'$97%$'+ ' = '+str(round(10**c97Int,2))+'*'+r'$N$'+'$^{'+str(round(c97Coef,2))+'}$', fontsize=fs, color='Crimson')
plt.text(1.35, 3.1, r'$95%$'+ ' = '+str(round(10**c95Int,2))+'*'+r'$N$'+'$^{'+str(round(c95Coef,2))+'}$', fontsize=fs, color='0.3')
plt.text(1.35, 2.7, r'$R^2$' + '=' +str(R2), fontsize=fs-1, color='k')
plt.xlabel('Number of reads, '+ '$log$'+r'$_{10}$', fontsize=fs)
plt.ylabel(metric, fontsize=fs)
plt.tick_params(axis='both', which='major', labelsize=fs-3)
plt.subplots_adjust(wspace=0.4, hspace=0.4)
if Ones =='N':
plt.savefig(mydir+'figs/appendix/PercentCutoff/PercentCutoff_NoMicrobe1s.png', dpi=600, bbox_inches = "tight")
elif Ones =='Y':
plt.savefig(mydir+'figs/appendix/PercentCutoff/PercentCutoff.png', dpi=600, bbox_inches = "tight")
#plt.show()
plt.close()
return
