def plotMeasurementOfParameter(data, parameter, value, agg="mean", color=None, label=None):
agg_func = {"mean": np.mean, "median": np.median, "mode": get_mode_stats, "logmode": get_mode_stats_log}[agg]
agg_func_err = {"mean": lambda x: np.std(x)/len(x), "median": lambda x: bootstrap_error(x, np.median),
"mode": lambda x: bootstrap_error(x, get_mode_stats, repetitions=100), "logmode": lambda x: bootstrap_error(x, get_mode_stats_log, repetitions=100)}[agg]
if 0:
plt.violinplot([meas_data[value].values for measurement_id, meas_data in data.groupby(parameter)], positions=np.log10(data[parameter].unique()))
#data.groupby(parameter)[value].boxplot()
if 1:
#agg = data.groupby([parameter, "measurement_id"])[value].agg(agg_func).reset_index()
# agg_err = data.groupby([parameter, "measurement_id"])[value].agg(get_mode_stats_err).reset_index()
# agg = data.groupby([parameter, "measurement_id"])[value].mean().reset_index()
# agg_err = data.groupby([parameter, "measurement_id"])[value].sem().reset_index()
agg_mean = data.groupby(parameter)[value].agg(agg_func)
agg_err_mean = data.groupby(parameter)[value].agg(agg_func_err) # agg(lambda x: 0.5*np.sqrt((x**2).sum()))
#l = plt.errorbar(np.log10(agg_mean.index), agg_mean.values, agg_err_mean.values, capsize=3, color=color, zorder=2, label=label)
l = plt.errorbar(agg_mean.index, agg_mean.values, agg_err_mean.values, capsize=3, color=color, zorder=2, label=label)
else:
agg = data.groupby([parameter, "measurement_id"])[value].agg(agg_func).reset_index()
#agg_err = data.groupby([parameter, "measurement_id"])[value].agg(get_mode_stats_err).reset_index()
#agg = data.groupby([parameter, "measurement_id"])[value].mean().reset_index()
#agg_err = data.groupby([parameter, "measurement_id"])[value].sem().reset_index()
agg_mean = agg.groupby(parameter)[value].mean()
agg_err_mean = agg.groupby(parameter)[value].sem()#agg(lambda x: 0.5*np.sqrt((x**2).sum()))
l = plt.errorbar(agg_mean.index, agg_mean.values, agg_err_mean.values, capsize=3, color=color, zorder=2, label=label)
for measurement_id, meas_data in agg.groupby("measurement_id"):
plt.plot(meas_data[parameter], meas_data[value], "o", ms=3, color=l[0].get_color(), alpha=0.5)
plt.plot(meas_data[parameter], meas_data[value], "-", ms=3, color="gray", alpha=0.5)
#plt.text(meas_data[parameter].values[0], meas_data[value].values[0], measurement_id)
plt.xlabel(parameter)
plt.ylabel(value)
def get_mode_stats(x):
from deformationcytometer.evaluation.helper_functions import bootstrap_error
from scipy import stats
x = np.array(x)
x = x[~np.isnan(x)]
def get_mode(x):
kde = stats.gaussian_kde(x)
return x[np.argmax(kde(x))]
mode = get_mode(x)
err = bootstrap_error(x, get_mode, repetitions=10)
def string(x):
if x > 1:
return str(round(x))
else:
return str(round(x, 2))
plt.text(0.5,
1,
string(mode) + "$\pm$" + string(err),
transform=plt.gca().transAxes,
ha="center",
va="top")
return mode, err, len(x)
def plotBinnedData(x, y, bins, bin_func=np.median, error_func=None, color="black", **kwargs):
x = np.asarray(x)
y = np.asarray(y)
strain_av = []
stress_av = []
strain_err = []
for i in range(len(bins) - 1):
index = (bins[i] < x) & (x < bins[i + 1])
yy = y[index]
yy = yy[~np.isnan(yy)]
if len(yy) == 0:
continue
strain_av.append(bin_func(yy))
# yy = yy[yy>0]
# strain_err.append(np.std(np.log(yy)) / np.sqrt(len(yy)))
if error_func is None:
strain_err.append(bootstrap_error(yy, bin_func)) # np.quantile(yy, [0.25, 0.75]))
elif error_func == "quantiles":
strain_err.append(np.abs(np.quantile(yy, [0.25, 0.75])-bin_func(yy))) # np.quantile(yy, [0.25, 0.75]))
stress_av.append(np.median(x[index]))
plot_kwargs = dict(marker='s', mfc="white", mec=color, ms=7, mew=1, lw=0, ecolor='black', elinewidth=1, capsize=3)
plot_kwargs.update(kwargs)
plt.errorbar(stress_av, strain_av, yerr=np.array(strain_err).T, **plot_kwargs)
x, y = np.array(stress_av), np.array(strain_av)
index = ~np.isnan(x) & ~np.isnan(y)
x = x[index]
y = y[index]
return x, y
def get_mode_stats_err(x):
from scipy import stats
from deformationcytometer.evaluation.helper_functions import bootstrap_error
x = np.array(x)
x = x[~np.isnan(x)&(x>0)]
def get_mode(x):
x = np.log(x)
kde = stats.gaussian_kde(x)
return np.exp(x[np.argmax(kde(x))])
#mode = get_mode(x)
err = bootstrap_error(x, get_mode, repetitions=2)
return err