def slm_plot_scaleloc(fitted_model,
ax=None,
scolor='C0',
lcolor='C1',
lw=2,
lowess=True,
annotations=3):
"""Produce scale-location plot."""
resids = np.sqrt(
np.abs(fitted_model.get_influence().resid_studentized_internal))
values = fitted_model.fittedvalues
ax = sns.scatterplot(values, resids, ax=ax, color=scolor)
if lowess:
ax.plot(*utils.lowess(values, resids), color=lcolor, lw=lw)
if annotations:
idxs = pd.Series(resids).nlargest(annotations).index
for idx in idxs:
value = values[idx]
resid = resids[idx]
ax.annotate(values.index[idx], (value, resid))
ax.set_title('Scale-Location')
ax.set_xlabel('Fitted Values')
ax.set_ylabel('$\sqrt{|\mathrm{Standardised}\; \mathrm{Residuals}|}$')
return ax
def glm_plot_scaleloc(fitted_model,
ax=None,
scolor='C0',
lcolor='C1',
lw=2,
lowess=True,
annotations=3):
"""Produce scale-location plot."""
resids = np.sqrt(np.abs(fitted_model.resid_deviance))
values = fitted_model.fittedvalues
ax = sns.scatterplot(values, resids, ax=ax, color=scolor)
if lowess:
ax.plot(*utils.lowess(values, resids), color=lcolor, lw=lw)
if annotations:
idxs = pd.Series(resids).nlargest(annotations).index
for idx in idxs:
value = values[idx]
resid = resids[idx]
ax.annotate(values.index[idx], (value, resid))
ax.set_title('Scale-Location')
ax.set_xlabel('Predicted Values')
ax.set_ylabel(
'$\sqrt{|\mathrm{Std.}\; \mathrm{Deviance}\; \mathrm{Resid.}|}$')
#_, y_high = ax.get_ylim()
#ax.set_ylim((-0.01, y_high))
return ax
def slm_plot_resid(fitted_model,
ax=None,
scolor='C0',
lcolor='C1',
lw=2,
lowess=True,
annotations=3):
"""Plot residuals versus fitted values."""
values = fitted_model.fittedvalues
resids = fitted_model.resid
ax = sns.scatterplot(values, resids, color=scolor, ax=ax)
ax.axhline(0, color=scolor, alpha=0.5)
if lowess:
ax.plot(*utils.lowess(values, resids), color=lcolor, lw=lw)
if annotations:
idxs = resids.abs().nlargest(annotations).index
for idx in idxs:
value = values[idx].max()
resid = resids[idx].max()
ax.annotate(idx, (value, resid))
ax.set_title('Residuals vs Fitted')
ax.set_xlabel('Fitted Values')
ax.set_ylabel('Residuals')
return ax
def glm_plot_leverage(fitted_model,
ax=None,
scolor='C0',
lcolor='C1',
ccolor='C2',
lw=2,
lowess=True,
cook=True,
legend=True,
annotations=3):
"""Produce leverage plot."""
influence = fitted_model.get_influence()
resids = fitted_model.resid_pearson
values = influence.hat_matrix_diag
cooks = influence.cooks_distance[0]
ax = sns.scatterplot(values, resids, ax=ax, color=scolor)
ax.axhline(0, color=scolor, alpha=0.5)
if lowess:
ax.plot(*utils.lowess(values, resids), color=lcolor, lw=lw)
if annotations:
idxs = pd.Series(cooks).nlargest(annotations).index
for idx in idxs:
value = values[idx]
resid = resids[idx]
ax.annotate(fitted_model.fittedvalues.index[idx], (value, resid))
if cook:
xlim = ax.get_xlim()
ylim = ax.get_ylim()
p = fitted_model.params.size
x = np.linspace(*xlim, 100)
y = np.sqrt((0.5 * p * (1 - x)) / x)
ax.plot(x, y, color=ccolor, lw=lw, alpha=0.8, label="Cook's Distance")
ax.set_xlim(xlim)
ax.set_ylim(ylim)
if legend:
ax.legend()
ax.set_title('Residuals vs Leverage')
ax.set_xlabel('Leverage')
ax.set_ylabel('Std. Pearson Resid.')
return ax
def plot_fit(fitted_model,
column,
data=None,
ax=None,
points=100,
scolor='C0',
fcolor='C1',
pcolor='C2',
lcolor='C7',
cialpha=0.3,
pialpha=0.2,
lalpha=0.8,
lw=2,
show_ci=True,
show_pi=False,
lowess=False,
legend=False):
"""Make a scatter plot and overlay fit result.
Make a scatter plot of the response versus a specified predictor and
overlay the fit result. The distributions of the other predictors are
marginalised out, ie. they are set to the mean values of their respective
distributions.
Returns the matplot lib axis object the plot was drawn on.
"""
model = fitted_model.model
if data is None:
data = pd.DataFrame(model.exog, columns=model.exog_names)
xs = utils.marginalised_range(column, data, points=points)
pred = fitted_model.get_prediction(xs).summary_frame()
x = xs[column]
y = pred['mean']
cil = pred['mean_ci_lower']
ciu = pred['mean_ci_upper']
slabel = None
if legend:
slabel = 'data'
ax = sns.scatterplot(x=column,
y=model.endog,
data=data,
ax=ax,
color=scolor,
label=slabel)
ax.set_title(f'Fit vs {column}')
ax.set_ylabel(model.endog_names)
ax.plot(x, y, color=fcolor, lw=lw, label='fit')
if show_ci:
ax.fill_between(x,
cil,
ciu,
color=fcolor,
alpha=cialpha,
label='conf. int.')
if show_pi:
try:
pil = pred['obs_ci_lower']
piu = pred['obs_ci_upper']
ax.fill_between(x,
pil,
piu,
color=pcolor,
alpha=pialpha,
label='pred. int.')
except KeyError:
pass
if lowess:
ax.plot(*utils.lowess(data[column], model.endog),
color=lcolor,
lw=lw,
alpha=lalpha,
label='lowess')
if legend:
ax.legend()
return ax