def descStats(data):
"""
Compute descriptive statistics of data
"""
dataList = list(data)
logDataList = list(N.log10(dataList))
desc = dict()
if len(dataList) == 0:
desc['mean'] = 0
desc['median'] = 0
desc['logMean'] = 0
desc['logMedian'] = 0
elif len(dataList) < 2:
desc['mean'] = dataList[0]
desc['median'] = dataList[0]
desc['logMean'] = logDataList[0]
desc['logMedian'] = logDataList[0]
else:
desc['mean'] = mean(dataList)
desc['median'] = median(dataList)
desc['logMean'] = mean(logDataList)
desc['logMedian'] = median(logDataList)
if len(dataList) < 3:
desc['stdev'] = 0
desc['sterr'] = 0
desc['logStdev'] = 0
desc['logSterr'] = 0
else:
desc['stdev'] = std(dataList)
desc['sterr'] = stderr(dataList)
desc['logStdev'] = std(logDataList)
desc['logSterr'] = stderr(logDataList)
return desc
def test_2d(self):
a = [[1.0, 2.0, 3.0], [2.0, 4.0, 6.0], [8.0, 12.0, 7.0]]
b1 = array(
(3.7859388972001824, 5.2915026221291814, 2.0816659994661335))
b2 = array((1.0, 2.0, 2.64575131106))
assert_array_almost_equal(stats.std(a), b1, 11)
assert_array_almost_equal(stats.std(a, axis=0), b1, 11)
assert_array_almost_equal(stats.std(a, axis=1), b2, 11)
def test_2d(self):
a = [[1.0, 2.0, 3.0],
[2.0, 4.0, 6.0],
[8.0, 12.0, 7.0]]
b1 = array((3.7859388972001824, 5.2915026221291814,
2.0816659994661335))
b2 = array((1.0,2.0,2.64575131106))
assert_array_almost_equal(stats.std(a),b1,11)
assert_array_almost_equal(stats.std(a,axis=0),b1,11)
assert_array_almost_equal(stats.std(a,axis=1),b2,11)
def errorfill(x, y, yerr = None, ax = None, color = None, err_type = 'sem', **kwargs): ax, fig = axis_check(ax) if color is None: color = ax._get_lines.color_cycle.next() if len(y.shape)==1: y_mean = y else: y_mean = y.mean(0) if yerr is None: if err_type == 'sem': yerr = 0.5*st.sem(y, 0) elif err_type == 'std': yerr = 0.5*st.std(y, 0) x_hi = y_mean + yerr x_lo = y_mean - yerr l = ax.plot(x, y_mean, color = color, **kwargs) l_up = ax.plot(x, y_mean+yerr, color = color, alpha = 0.2) l_lo = ax.plot(x, y_mean-yerr, color = color, alpha = 0.2) ax.fill_between(x, x_hi, x_lo, alpha = 0.2, color = color) return ax
def gStats(self, missingValue=0.0):
"""dict of {geneID: (min,max,mean,median,std,stderr,
Shapiro-Wilk(w,p),normaltest_chisq (D'Agostino and Pearson),...}
"""
import scipy as S
import scipy.stats as SS
rv = {}
for k, v in self.items():
# print k,v
va = S.array(self.gValues(k, missingValue))
try:
normaltest = SS.normaltest(va)
except:
normaltest = None
try:
shapiro = SS.shapiro(va)
except:
shapiro = None
try:
rv[k] = (va.min(), va.max(), va.mean(), SS.median(va), SS.std(va), SS.stderr(va), normaltest, shapiro)
except:
print k, va
raise
return rv
def _remove_outliers(data):
return data
if len(data) < 2:
return data
else:
lmean = mean(data)
limstdev = 3 * std(data)
return [item for item in data if abs(item - lmean) < limstdev ]
def reindex(self):
period = self.series[-self.period:]
last = self.series[-1]
try:
dev = std(period)
dev *= self.dev_factor
dev += last
except (TypeError, ZeroDivisionError, ):
dev = None
self.append(dev)
def reindex(self):
periods = self.periods
period = self.series[-periods:]
vol = None
if len(period) == periods:
try:
vol = std(period) / mean(period)
vol *= 100
except TypeError:
pass
self.append(vol)
def __init__(self, samples):
self.samples = numpy.asarray(samples)
self.N = len(samples)
self.median = stats.median(samples)
self.min = numpy.amin(samples)
self.max = numpy.amax(samples)
self.mean = stats.mean(samples)
self.std = stats.std(samples)
self.var = self.std**2.
self.skew = stats.skew(samples)
self.kurtosis = stats.kurtosis(samples)
self.range = self.max - self.min
def _getcs(self, data):
if self.c:
c = data[self.c]
#print data, c
sigma, mu = std(c), mean(c)
c = N.clip(c, mu - 2 * sigma, mu + 2 * sigma)
c = (max(c) - c) / (max(c) - min(c))
else:
c = 'b'
if self.s:
s = data[self.s]
else:
s = 10.0
return c, s
def _getcs(self, data):
if self.c:
c = data[self.c]
#print data, c
sigma, mu = std(c), mean(c)
c = N.clip(c, mu - 2*sigma, mu + 2*sigma)
c = (max(c) - c)/(max(c) - min(c))
else:
c = 'b'
if self.s:
s = data[self.s]
else:
s = 10.0
return c, s
def confidence(samples, confidence_level):
"""This function determines the confidence interval for a given set of samples,
as well as the mean, the standard deviation, and the size of the confidence
interval as a percentage of the mean.
From javastats by Andy Georges.
"""
mean = stats.mean(samples)
sdev = stats.std(samples)
n = len(samples)
df = n - 1
t = distributions.t.ppf((1+confidence_level)/2.0, df)
interval = (interval_low, interval_high) = ( mean - t * sdev / math.sqrt(n) , mean + t * sdev / math.sqrt(n) )
interval_size = interval_high - interval_low
interval_percentage = interval_size / mean * 100.0
return (interval, mean, sdev, interval_percentage)
def skewsField(sample, field):
"""
Checks whether the value of field in the passed in sample is significantly different from the
value of field for the rest of the samples under consideration.
"""
savedSamples = samples.sampleList[:]
samples.sampleList.remove(sample)
try:
flds = samples.getAllFlds(field)
mean = stats.mean(flds)
stddev = stats.std(flds)
val = sample[field]
if stddev == 0:
devs = 0
else:
devs = abs(val - mean) / stddev
finally:
#we should be fixing the sample list even when I crash!
samples.sampleList = savedSamples
if len(samples.sampleList) < 3:
qual = confidence.Validity.plaus
elif len(samples.sampleList) < 6:
qual = confidence.Validity.prob
else:
qual = confidence.Validity.sound
conf = __getConfidence((.5, 1, 2, 3, 5), devs, qual)
samples.sampleList.sort(key=lambda x: samples.extractField(x, field))
plot = __getPlot('id', field)
plot.plotLine(0, mean)
plot.plotLine(0, mean-stddev)
plot.plotLine(0, mean+stddev)
plot.plotLine(0, sample[field])
return SimResult(conf, str(sample) + " has a different " + field + " from other samples",
str(sample) + "'s value for " + field + ' is ' + str(devs) +
' standard deviations from the mean', plot)
def main():
if( len(sys.argv) != 3 ):
printUsage()
executable = sys.argv[1]
no_times = int(sys.argv[2])
cmd = "time ./%s" % executable
times = zeros(no_times, 'f');
for i in range(no_times):
t = time(); os.system(cmd); t_store = time() - t;
print "t_store= ", t_store
times[i] = t_store;
print "Mean time for operation = ", mean(times)
print "Standard deviation for operation = ", std(times)
def fit_gaussians(data, initial_params, errs, profnm):
numparams = len(initial_params)
numgaussians = (len(initial_params)-1)/3
# Generate the parameter structure
parinfo = []
params0 = []
for ii in range(numparams):
params0.append(initial_params[ii])
parinfo.append({'value':initial_params[ii], 'fixed':0,
'limited':[0,0], 'limits':[0.,0.]})
other_args = {'data':data, 'errs':errs}
# Now fit it
mpfit_out = mpfit.mpfit(fit_function, params0, functkw=other_args,
parinfo=parinfo, quiet=1)
fit_params = mpfit_out.params
fit_errs = mpfit_out.perror
# degrees of freedom
dof = len(data) - len(fit_params)
# chi-squared for the model fit
chi_sq = mpfit_out.fnorm
print "------------------------------------------------------------------"
print "Multi-Gaussian Fit by pygaussfit.py of '%s'"%profnm
print "------------------------------------------------------------------"
print "mpfit status:", mpfit_out.status
print "gaussians:", numgaussians
print "DOF:", dof
print "chi-sq: %.2f" % chi_sq
print "reduced chi-sq: %.2f" % (chi_sq/dof)
residuals = data - gen_gaussians(fit_params, len(data))
print "residuals mean: %.3g"%mean(residuals)
print "residuals stdev: %.3g"%std(residuals)
print "--------------------------------------"
print " const = %.5f +/- %.5f" % (fit_params[0], fit_errs[0])
for ii in range(numgaussians):
print " phas%d = %.5f +/- %.5f" % (ii+1, fit_params[1+ii*3], fit_errs[1+ii*3])
print " fwhm%d = %.5f +/- %.5f" % (ii+1, fit_params[2+ii*3], fit_errs[2+ii*3])
print " ampl%d = %.5f +/- %.5f" % (ii+1, fit_params[3+ii*3], fit_errs[3+ii*3])
print "--------------------------------------"
return fit_params, fit_errs, chi_sq, dof
def histo_plotter(x, SPECS):
# the histogram of the data
n, bins, patches = hist(x, 50, normed=0)
setp(patches, 'facecolor', 'g', 'alpha', 0.75)
for i in range(len(SPECS)):
out_of_spec = 0
for j in range(len(bins)):
if (bins[j] <= SPECS[i]):
setp(patches[j], 'facecolor', 'r', 'alpha', 0.75)
out_of_spec = out_of_spec + n[j]
out_summary="#of Samples Below "+str(SPECS[i])+" :"+str(out_of_spec)+\
"\n From "+str(len(x))+" Samples "
fp3.write(out_summary)
#print patches
# add a 'best fit' line
mu = stats.mean(x)
sigma = stats.std(x)
maxfreq = max(n)
minval = min(x)
out_summary2 = "Minimum Value: " + fix(minval, 3)
fp3.write(out_summary2)
# print x, bins, n, mu, sigma
y = normpdf(bins, mu, sigma)
l = plot(bins, y, 'r--')
#y = normpdf( bins)
#l = plot(bins, n, 'r--')
setp(l, 'linewidth', 1)
xlabel('Clearance')
ylabel('Count')
title(r'$\rm{Histogram\ of\ Clearance}$')
axis([bins[0], bins[49], 0.0, maxfreq])
text(.01 + bins[0], .9 * maxfreq, out_summary, color='r')
text(.01 + bins[0], .8 * maxfreq, out_summary2, color='b')
grid(True)
#savefig('histogram_demo',dpi=72)
show()
def test_stdX(self):
y = stats.std(X)
assert_almost_equal(y, 2.738612788)
def test_std(self):
y = stats.std(self.testcase)
assert_approx_equal(y,1.290994449)
def test_stdHUGE(self):
y = stats.std(HUGE)
assert_approx_equal(y, 2.738612788e12)
def test_basic(self):
a = [3,4,5,10,-3,-5,6]
b = [3,4,5,10,-3,-5,-6]
assert_almost_equal(stats.std(a),5.2098807225172772,11)
assert_almost_equal(stats.std(b),5.9281411203561225,11)
def test_stdBIG(self):
y = stats.std(BIG)
assert_almost_equal(y, 2.738612788)
def calculate_mean_std(metric):
values = list(perf[metric].values())
perf['metrics'][metric]['mean'] = np.mean(values)
perf['metrics'][metric]['std'] = std(values)
def cross_validation(X, y, n_trials=5, trial_splits=None, fname=None):
"""Perform model selection via 5-fold cross validation"""
# filter samples with no annotations
del_rid = np.where(y.sum(axis=1) == 0)[0]
y = np.delete(y, del_rid, axis=0)
X = np.delete(X, del_rid, axis=0)
# range of hyperparameters
C_range = 10.**np.arange(-1, 3)
gamma_range = 10.**np.arange(-3, 1)
# pre-generating kernels
print("### Pregenerating kernels...")
K_rbf = {}
for gamma in gamma_range:
K_rbf[gamma] = rbf_kernel(X, gamma=gamma)
print("### Done.")
# performance measures
perf = dict()
pr_micro = []
pr_macro = []
fmax = []
acc = []
if trial_splits is None:
# shuffle and split training and test sets
trials = ShuffleSplit(n_splits=n_trials,
test_size=0.2,
random_state=None)
ss = trials.split(X)
trial_splits = []
for train_idx, test_idx in ss:
trial_splits.append((train_idx, test_idx))
it = 0
for jj in range(0, n_trials):
train_idx = trial_splits[jj][0]
test_idx = trial_splits[jj][1]
it += 1
y_train = y[train_idx]
y_test = y[test_idx]
print("### [Trial %d] Perfom cross validation...." % (it))
print("Train samples=%d; #Test samples=%d" %
(y_train.shape[0], y_test.shape[0]))
# setup for neasted cross-validation
splits = ml_split(y_train)
# parameter fitting
C_opt = None
gamma_opt = None
max_aupr = 0
for C in C_range:
for gamma in gamma_range:
# Multi-label classification
cv_results = []
for train, valid in splits:
clf = OneVsRestClassifier(svm.SVC(C=C,
kernel='precomputed',
probability=False),
n_jobs=-1)
K_train = K_rbf[gamma][
train_idx[train], :][:, train_idx[train]]
K_valid = K_rbf[gamma][
train_idx[valid], :][:, train_idx[train]]
y_train_t = y_train[train]
y_train_v = y_train[valid]
y_score_valid = np.zeros(y_train_v.shape, dtype=float)
y_pred_valid = np.zeros_like(y_train_v)
idx = np.where(y_train_t.sum(axis=0) > 0)[0]
clf.fit(K_train, y_train_t[:, idx])
y_score_valid[:, idx] = clf.decision_function(K_valid)
y_pred_valid[:, idx] = clf.predict(K_valid)
perf_cv = evaluate_performance(y_train_v, y_score_valid,
y_pred_valid)
cv_results.append(perf_cv['m-aupr'])
cv_aupr = np.median(cv_results)
print("### gamma = %0.3f, C = %0.3f, AUPR = %0.3f" %
(gamma, C, cv_aupr))
if cv_aupr > max_aupr:
C_opt = C
gamma_opt = gamma
max_aupr = cv_aupr
print("### Optimal parameters: ")
print("C_opt = %0.3f, gamma_opt = %0.3f" % (C_opt, gamma_opt))
print("### Train dataset: AUPR = %0.3f" % (max_aupr))
print("### Using full training data...")
clf = OneVsRestClassifier(svm.SVC(C=C_opt,
kernel='precomputed',
probability=False),
n_jobs=-1)
y_score = np.zeros(y_test.shape, dtype=float)
y_pred = np.zeros_like(y_test)
idx = np.where(y_train.sum(axis=0) > 0)[0]
clf.fit(K_rbf[gamma_opt][train_idx, :][:, train_idx], y_train[:, idx])
# Compute performance on test set
y_score[:, idx] = clf.decision_function(
K_rbf[gamma_opt][test_idx, :][:, train_idx])
y_pred[:, idx] = clf.predict(K_rbf[gamma_opt][test_idx, :][:,
train_idx])
perf_trial = evaluate_performance(y_test, y_score, y_pred)
pr_micro.append(perf_trial['m-aupr'])
pr_macro.append(perf_trial['M-aupr'])
fmax.append(perf_trial['F1'])
acc.append(perf_trial['acc'])
print(
"### Test dataset: AUPR['micro'] = %0.3f, AUPR['macro'] = %0.3f, F1 = %0.3f, Acc = %0.3f"
% (perf_trial['m-aupr'], perf_trial['M-aupr'], perf_trial['F1'],
perf_trial['acc']))
perf['m-aupr_avg'] = np.mean(pr_micro)
perf['m-aupr_std'] = std(pr_micro)
perf['M-aupr_avg'] = np.mean(pr_macro)
perf['M-aupr_std'] = std(pr_macro)
perf['F1_avg'] = np.mean(fmax)
perf['F1_std'] = std(fmax)
perf['acc_avg'] = np.mean(acc)
perf['acc_std'] = std(acc)
if fname is not None:
fout = open(fname, 'w')
fout.write("aupr[micro], aupr[macro], F_max, accuracy\n")
for ii in range(0, n_trials):
fout.write(pr_micro[ii], pr_macro[ii], fmax[ii], acc[ii])
fout.close()
return perf
def test_stdTINY(self):
y = stats.std(TINY)
assert_almost_equal(y, 0.0)
def test_stdBIG(self):
y = stats.std(BIG)
assert_almost_equal(y, 2.738612788)
def temporal_holdout(X,
y,
indx,
bootstrap,
fname,
goterms=None,
go_fname=None):
"""Perform temporal holdout validation"""
X_train = X[indx['train'].tolist()]
X_test = X[indx['test'].tolist()]
X_valid = X[indx['valid'].tolist()]
y_train = y['train'].tolist()
y_test = y['test'].tolist()
y_valid = y['valid'].tolist()
if goterms is not None:
goterms = goterms['terms'].tolist()
# range of hyperparameters
C_range = 10.**np.arange(-1, 3)
gamma_range = 10.**np.arange(-3, 1)
# pre-generating kernels
print("### Pregenerating kernels...")
K_rbf_train = {}
K_rbf_test = {}
K_rbf_valid = {}
for gamma in gamma_range:
K_rbf_train[gamma] = rbf_kernel(X_train, gamma=gamma)
K_rbf_test[gamma] = rbf_kernel(X_test, X_train, gamma=gamma)
K_rbf_valid[gamma] = rbf_kernel(X_valid, X_train, gamma=gamma)
print("### Done.")
print("Train samples=%d; #Test samples=%d" %
(y_train.shape[0], y_test.shape[0]))
# parameter fitting
C_opt = None
gamma_opt = None
max_aupr = 0
for C in C_range:
for gamma in gamma_range:
# Multi-label classification
clf = OneVsRestClassifier(svm.SVC(C=C,
kernel='precomputed',
probability=False),
n_jobs=-1)
clf.fit(K_rbf_train[gamma], y_train)
y_score_valid = clf.decision_function(K_rbf_valid[gamma])
y_pred_valid = clf.predict(K_rbf_valid[gamma])
perf = evaluate_performance(y_valid, y_score_valid, y_pred_valid)
micro_aupr = perf['m-aupr']
print("### gamma = %0.3f, C = %0.3f, AUPR = %0.3f" %
(gamma, C, micro_aupr))
if micro_aupr > max_aupr:
C_opt = C
gamma_opt = gamma
max_aupr = micro_aupr
print("### Optimal parameters: ")
print("C_opt = %0.3f, gamma_opt = %0.3f" % (C_opt, gamma_opt))
print("### Train dataset: AUPR = %0.3f" % (max_aupr))
print("### Computing performance on test dataset...")
clf = OneVsRestClassifier(svm.SVC(C=C_opt,
kernel='precomputed',
probability=False),
n_jobs=-1)
clf.fit(K_rbf_train[gamma_opt], y_train)
# Compute performance on test set
y_score = clf.decision_function(K_rbf_test[gamma_opt])
y_pred = clf.predict(K_rbf_test[gamma_opt])
# performance measures for bootstrapping
perf = dict()
pr_micro = []
pr_macro = []
fmax = []
acc = []
# individual goterms
pr_goterms = {}
for i in range(0, len(goterms)):
pr_goterms[goterms[i]] = []
for ind in bootstrap:
perf_ind = evaluate_performance(y_test[ind], y_score[ind], y_pred[ind])
pr_micro.append(perf_ind['m-aupr'])
pr_macro.append(perf_ind['M-aupr'])
fmax.append(perf_ind['F1'])
acc.append(perf_ind['acc'])
for i in range(0, len(goterms)):
pr_goterms[goterms[i]].append(perf_ind[i])
perf['m-aupr_avg'] = np.mean(pr_micro)
perf['m-aupr_std'] = std(pr_micro)
perf['M-aupr_avg'] = np.mean(pr_macro)
perf['M-aupr_std'] = std(pr_macro)
perf['F1_avg'] = np.mean(fmax)
perf['F1_std'] = std(fmax)
perf['acc_avg'] = np.mean(acc)
perf['acc_std'] = std(acc)
# trials
fout = open(fname, 'w')
fout.write("aupr[micro], aupr[macro], F_max, accuracy\n")
for it in range(0, len(bootstrap)):
fout.write(pr_micro[it], pr_macro[it], fmax[it], acc[it], "\n")
fout.close()
# write performance on individual GO terms
if go_fname is not None:
fout = open(go_fname, 'wb')
print >> fout, "GO_id, AUPRs"
for i in range(0, len(goterms)):
print >> fout, goterms[i], sum(y_train[:, i]) / float(
y_train.shape[0]),
for pr in pr_goterms[goterms[i]]:
print >> fout, pr,
print >> fout
fout.close()
return perf
def test_stdZERO(self):
y = stats.std(ZERO)
assert_almost_equal(y, 0.0)
def test_stdX(self):
y = stats.std(X)
assert_almost_equal(y, 2.738612788)
def test_std(self):
y = stats.std(self.testcase)
assert_approx_equal(y, 1.290994449)
def test_stdROUND(self):
y = stats.std(ROUND)
assert_approx_equal(y, 2.738612788)
def test_nanstd_some(self):
"""Check nanstd when some values only are nan."""
s = stats.nanstd(self.Xsome)
assert_approx_equal(s, stats.std(self.Xsomet))
def table_fit_many(table, ff, fit_bounds):
if fit_bounds is not None:
if (stats.std(fits, axis=0) > stats.std(fit.boot)).any():
warnings.warn("BAD")
def test_nanstd_none(self):
"""Check nanstd when no values are nan."""
s = stats.nanstd(self.X)
assert_approx_equal(s, stats.std(self.X))
cc.cycle(dt,fixnuc = 0)
if run % 5 == 0:
cc.display(dt)
length_proj = S.array(cc.length_proj)
trajs_theta = S.array(cc.trajs_theta)
trans12 = int(cc.temps_1 * 0.5)
trans23 = int(cc.temps_2 * 0.5)
if len(length_proj) > trans23:
longueur_moyenne.append(stats.mean(length_proj[trans12:trans23]))
angle_moyen.append(stats.mean(trajs_theta[trans12:trans23]))
angle_maxi.append(reduce(maximum,abs(trajs_theta[trans12:trans23])))
angle_onset.append(abs(trajs_theta[trans23]))
ecartype.append(stats.std(trajs_theta[trans12:trans23]))
duree1.append(cc.temps_1)
duree2.append(cc.temps_2 - cc.temps_1)
fname = 'cell'+str(run)
cc.logging(fname,dt)
else :
print 'discarded cell # %d ' % run
del cc
angle_moyen = S.array(angle_moyen)
longueur_moyenne= S.array(longueur_moyenne)
ecartype = S.array(ecartype)
duree2 = S.array(duree2)
def test_basic(self):
a = [3, 4, 5, 10, -3, -5, 6]
b = [3, 4, 5, 10, -3, -5, -6]
assert_almost_equal(stats.std(a), 5.2098807225172772, 11)
assert_almost_equal(stats.std(b), 5.9281411203561225, 11)
def zscore(line):
'''replacement for scipy.stats.z, but uses N-1 degrees of freedom,
instead of N degrees of freedom'''
return (line - np.mean(line))/stats.std(line)
from scipy import stats import sys ephemerides = ReadEphemeridesLog(sys.argv[1]) comps = [] for e in ephemerides: c = ComparePysolarToUSNO(e) comps.append(c) az_errors = [c.az_error for c in comps] alt_errors = [c.alt_error for c in comps] print '---------------------' print 'Azimuth stats' print 'Mean error: ' + str(stats.mean(az_errors)) print 'Std dev: ' + str(stats.std(az_errors)) print 'Min error: ' + str(stats.tmin(az_errors, None)) print 'Max error: ' + str(stats.tmax(az_errors, None)) print '----------------------' print 'Altitude stats' print 'Mean error: ' + str(stats.mean(alt_errors)) print 'Std dev: '+ str(stats.std(alt_errors)) print 'Min error: ' + str(stats.tmin(alt_errors, None)) print 'Max error: ' + str(stats.tmax(alt_errors, None)) WriteComparisonsToCSV(comps, 'pysolar_v_usno.csv')
def test_stdZERO(self):
y = stats.std(ZERO)
assert_almost_equal(y, 0.0)
def test_stdROUND(self):
y = stats.std(ROUND)
assert_approx_equal(y, 2.738612788)
def calculate_mean_std(performance, metric):
performance['metrics'][metric] = {}
values = list(performance[metric].values())
performance['metrics'][metric]['mean'] = np.mean(values)
performance['metrics'][metric]['std'] = std(values)
def test_nanstd_none(self):
"""Check nanstd when no values are nan."""
s = stats.nanstd(self.X)
assert_approx_equal(s, stats.std(self.X))
from scipy import stats
import sys
ephemerides = ReadEphemeridesLog(sys.argv[1])
comps = []
for e in ephemerides:
c = ComparePysolarToUSNO(e)
comps.append(c)
az_errors = [c.az_error for c in comps]
alt_errors = [c.alt_error for c in comps]
print '---------------------'
print 'Azimuth stats'
print 'Mean error: ' + str(stats.mean(az_errors))
print 'Std dev: ' + str(stats.std(az_errors))
print 'Min error: ' + str(stats.tmin(az_errors, None))
print 'Max error: ' + str(stats.tmax(az_errors, None))
print '----------------------'
print 'Altitude stats'
print 'Mean error: ' + str(stats.mean(alt_errors))
print 'Std dev: ' + str(stats.std(alt_errors))
print 'Min error: ' + str(stats.tmin(alt_errors, None))
print 'Max error: ' + str(stats.tmax(alt_errors, None))
WriteComparisonsToCSV(comps, 'pysolar_v_usno.csv')
def test_nanstd_some(self):
"""Check nanstd when some values only are nan."""
s = stats.nanstd(self.Xsome)
assert_approx_equal(s, stats.std(self.Xsomet))
def test_stdLITTLE(self):
y = stats.std(LITTLE)
assert_approx_equal(y, 2.738612788e-8)
def test_stdLITTLE(self):
y = stats.std(LITTLE)
assert_approx_equal(y, 2.738612788e-8)
def test_stdHUGE(self):
y = stats.std(HUGE)
assert_approx_equal(y, 2.738612788e12)
def test_stdTINY(self):
y = stats.std(TINY)
assert_almost_equal(y, 0.0)
def plotB5(B5s,vpns,clrs=None,exps=[1],suffix=''):
tt=0
for gg in range(len(B5s)):
B5=B5s[gg]; exp=exps[gg];vpn=vpns[gg]
titles=['Recalled Position','Selected Position','Selected Position']
posx=np.array([-6,-3,0,3,6,-6,-3,0],ndmin=2)
seln=~np.isnan(B5[:,0,0])
a=np.zeros(B5.shape[1]); a[:5]=1
b=np.zeros(B5.shape[1]);
if exp==1: b[5:10]=1
else: b[5:]=1
c=np.zeros(B5.shape[1]); c[10:]=1
sels=[a==1,b==1];kk=1
if exp==1: sels.append(c==1)
if clrs is None or len(clrs)!=B5.shape[0]: clrs=getColors(B5.shape[0])
#if gg==1: clrs=getColors(28)[:18]
#elif gg==2: clrs=getColors(28)[18:]
res=[]
for sel in sels:
objs=[]
v=[]
subplot(len(B5s),3,gg*3+kk);plt.grid(axis='y');kk+=1
if exp==1: drawCircularAgent(pos=(0,0),eyes=kk<4)
else: drawDartAgent(pos=(0,0))
plt.plot([-1,1],[0,0],color='#262626',label='_nolegend_',lw=0.5);
plt.plot([0,0],[-1,1],color='#262626',lw=0.5,label='_nolegend_')
dx=B5[:,sel,1]-posx[0,:sel.sum()]
dy=B5[:,sel,2]
mag=np.sqrt(np.power(dx,2)+np.power(dy,2))
if exp==1: phi=np.arctan2(dy,dx)-B5[:,sel,0]
else: phi=np.arctan2(dy,dx)-np.pi*B5[:,sel,0]/180
xn=np.cos(phi)*mag
yn=np.sin(phi)*mag
for i in range(B5.shape[0]):
plt.plot(xn[i,:],yn[i,:],'o',mfc=clrs[i],
label='_nolegend_',mew=0.15,alpha=0.5,ms=3,mec='k')
valid=np.sqrt(np.power(xn[i,:],2)+np.power(yn[i,:],2))<1.2
v.extend(valid.tolist())
if seln[i]: label=str(vpn[i])
else:label='_nolegend_'
plt.plot(np.median(xn[i,valid]),np.median(yn[i,valid]),'x',
label=label,mec=clrs[i],mew=1.5,ms=6,zorder=3)
xn[i,~valid]=np.nan
plt.gca().set_aspect('equal')
plt.xlim([-1,1]);plt.ylim([-1,1])
#if kk>2: plt.gca().set_yticklabels([])
if gg==0: plt.title(titles[kk-2],fontsize=12)
#print np.array(v).sum(),float(seln.sum()), len(v)/float(xn.shape[0])
res.append([xn])
xn=xn.flatten()[np.array(v)]
#if gg==1: xxn=np.copy(xn)
#elif gg==2: xn=np.concatenate([xxn,xn])
#xn=median(xn,1)
m=np.mean(xn)
plt.text(plt.xlim()[0]+0.1*(plt.xlim()[1]-plt.xlim()[0]),
plt.ylim()[1]-0.1*(plt.ylim()[1]-plt.ylim()[0]),
str(unichr(65+tt)),horizontalalignment='center',verticalalignment='center',
fontdict={'weight':'bold'},fontsize=12);tt+=1
plt.plot([m,m],[-1,1],'--g',color='gray',label='_nolegend_',zorder=-2)
sse=std(xn,bias=True)/xn.size**0.5
res[-1].extend([m,sse,xn.size])
er= sse* stats.t.ppf(0.975,xn.size)
er=[m-2*er,m+2*er]#[sap(xn,25),sap(xn,75)]
plt.gca().add_patch(plt.Rectangle([er[0],-5],er[1]-er[0],10,
color='k',zorder=-2,alpha=0.1))
print " %.3f CI [%.3f, %.3f]"%(m,er[0],er[1])
if False:#kk==2:
plt.legend(bbox_to_anchor=(0,0,1,1),loc=2,mode="expand",ncol=seln.sum()/2+1,
bbox_transform=plt.gcf().transFigure,frameon=False)
if exp==2:
x0=(1-3**0.5/2)*AR
for i in range(2):
plt.plot([x0,x0],[-1,1],':',color='gray')
return res
def least_squers_fit(x, y):
beta = numpy.corrcoef(x, y) * st.std(y) / st.std(x)
alpha = numpy.mean(y) - beta * numpy.mean(x)
return alpha, beta