def plot_bootstrap(self, fig=None):
"""Plot bootstrapped distribution."""
lower95CI, upper95CI = self.bootstrapped_CIs(alpha=0.05)
if fig is None:
fig, ax = plt.subplots(1, 1) #, figsize=(4,4))
else:
fig.clf()
ax = fig.add_subplot(1, 1, 1)
ax.hist(self.boot, 20, density=True)
xmin = self.bootstrap_mean - 4 * np.std(self.boot, ddof=1)
xmax = self.bootstrap_mean + 4 * np.std(self.boot, ddof=1)
increase = (xmax - xmin) / 100
x = np.arange(xmin, xmax, increase)
pdf = [
bs._pdf(x1, self.bootstrap_mean, np.std(self.boot, ddof=1))
for x1 in x
]
ax.plot(x, pdf, 'k', label='normal pdf')
ax.axvline(x=self.ratio, color='r', label='observed ratio')
ax.axvline(x=self.bootstrap_mean,
color='r',
linestyle="dashed",
label='bootstrapped ratio')
ax.axvline(x=lower95CI,
color='k',
linestyle="dashed",
label='2.5% limits')
ax.axvline(x=upper95CI, color='k', linestyle="dashed")
ax.set_ylabel("Frequency")
ax.set_xlabel('Ratio (mean A / mean B)')
ax.legend(bbox_to_anchor=(0., 1.02, 1., .102), loc=3, borderaxespad=0.)
plt.tight_layout()
return fig
def plot_rantest(self, fig=None):
"""Plot randomisation distribution."""
if fig is None:
fig, ax = plt.subplots(1, 1) #, figsize=(4,4))
else:
fig.clf()
ax = fig.add_subplot(1, 1, 1)
ax.hist(self.randiff, 20, density=True)
mu = np.mean(self.randiff)
sd = np.std(self.randiff, ddof=1)
xmin = mu - 4 * sd
xmax = mu + 4 * sd
increase = (xmax - xmin) / 100
x = np.arange(xmin, xmax, increase)
pdf = [bs._pdf(x1, mu, sd) for x1 in x]
ax.plot(x, pdf, 'k', label='normal pdf')
ax.axvline(x=self.dbar, color='r', label='observed difference')
ax.axvline(x=-self.dbar, color='r')
ax.axvline(x=self.lo95lim,
color='k',
linestyle='--',
label='2.5% limits')
ax.axvline(x=self.hi95lim, color='k', linestyle='--')
ax.set_xlabel('difference between means')
ax.set_ylabel('frequency')
ax.legend(bbox_to_anchor=(0., 1.02, 1., .102), loc=3, borderaxespad=0.)
plt.tight_layout()
return fig
def plot_bootstrapped_distributions(self, fig=None):
sample1 = Sample(self.A)
sample1.run_bootstrap(self.runs)
lower95CI1, upper95CI1 = sample1.bootstrapped_CIs(alpha=0.05)
sample2 = Sample(self.B)
sample2.run_bootstrap(self.runs)
lower95CI2, upper95CI2 = sample2.bootstrapped_CIs(alpha=0.05)
if fig is None:
fig = plt.Figure()
else:
fig.clf()
ax1 = fig.add_subplot(1, 2, 1)
ax1.hist(sample1.boot, bins=20, density=True)
mu = np.mean(sample1.boot)
sd = np.std(sample1.boot, ddof=1)
xmin = mu - 4 * sd
xmax = mu + 4 * sd
increase = (xmax - xmin) / 100
x = np.arange(xmin, xmax, increase)
pdf = [bs._pdf(x1, mu, sd) for x1 in x]
ax1.plot(x, pdf, 'k', label='normal pdf')
ax1.axvline(x=sample1.meanA, color='r', label='observed mean')
ax1.axvline(x=sample1.bootstrap_mean,
color='r',
linestyle="dashed",
label='bootstrapped mean')
ax1.axvline(x=lower95CI1,
color='k',
linestyle="dashed",
label='2.5% limits')
ax1.axvline(x=upper95CI1, color='k', linestyle="dashed")
ax1.set_ylabel("Frequency")
ax1.set_xlabel('Mean')
ax1.legend(bbox_to_anchor=(0., 1.02, 1., .102),
loc=3,
borderaxespad=0.)
ax2 = fig.add_subplot(1, 2, 2)
ax2.hist(sample2.boot, bins=20, density=True)
mu = np.mean(sample2.boot)
sd = np.std(sample2.boot, ddof=1)
xmin = mu - 4 * sd
xmax = mu + 4 * sd
increase = (xmax - xmin) / 100
x = np.arange(xmin, xmax, increase)
pdf = [bs._pdf(x1, mu, sd) for x1 in x]
ax2.plot(x, pdf, 'k', label='normal pdf')
ax2.axvline(x=sample2.meanA, color='r', label='observed mean')
ax2.axvline(x=sample2.bootstrap_mean,
color='r',
linestyle="dashed",
label='bootstrapped mean')
ax2.axvline(x=lower95CI2,
color='k',
linestyle="dashed",
label='2.5% limits')
ax2.axvline(x=upper95CI2, color='k', linestyle="dashed")
ax2.set_ylabel("Frequency")
ax2.set_xlabel('Mean')
ax2.legend(bbox_to_anchor=(0., 1.02, 1., .102),
loc=3,
borderaxespad=0.)
plt.tight_layout()
return fig