def plot_diagnostics(self, absolute=False):
plot_residuals(residual_array=self.r_mean, group_name='Overall mean', absolute=absolute)
if self.r_std is not None:
plot_residuals(residual_array=self.r_std, group_name='Overall std', absolute=True)
if self.r_mad is not None:
plot_residuals(residual_array=self.r_mad, group_name='Overall mad', absolute=True)
for k, v in self.pv_groups.items():
plot_residuals(residual_array=v.residuals, group_name=k, absolute=absolute)
def triangle_residual_plots(self,
radius,
x_axis='far_out',
y_axis='num_data',
exclude=None,
absolute=False):
"""
Plot all of the residuals based on some exclusion criteria for
number of data points that were used in the fitting and some radius
for smoothing over a window of neighboring data density / number predicting out.
Args:
radius: List[int]
absolute: (bool) plot mean residuals by value or absolute value
exclude: (int) exclude model fits with under this many data points
x_axis: (str) the x-axis variable to plot (one of far_out, num_data, or data_index)
y_axis: (str) the y-axis variable to plot (one of far_out, num_data, or data_index)
"""
smoothed_residuals = self.get_smoothed_residuals(radius=radius)
smoothed_residuals = smoothed_residuals.loc[
smoothed_residuals['num_data'] > exclude].copy()
for i, (k, v) in enumerate(self.pv_groups.items()):
residuals = self.all_residuals.loc[self.all_residuals[
self.col_group] == k].copy()
plot_residuals(residual_array=np.asarray(
residuals[[x_axis, y_axis, 'residual']]),
group_name=k,
absolute=absolute,
x_label=x_axis,
y_label=y_axis)
smooth = smoothed_residuals.loc[smoothed_residuals[self.col_group]
== k]
smooth_mean = np.asarray(smooth[[x_axis, y_axis, 'residual_mean']])
smooth_std = np.asarray(smooth[[x_axis, y_axis, 'residual_std']])
plot_residuals(residual_array=smooth_mean,
group_name=f'{k} smooth mean radius {radius}',
absolute=absolute,
x_label=x_axis,
y_label=y_axis)
plot_residuals(residual_array=smooth_std,
group_name=f'{k} smooth std radius {radius}',
absolute=True,
x_label=x_axis,
y_label=y_axis)
def plot_residuals(self, radius, absolute=False, exclude=5):
"""
Plot all of the residuals based on some exclusion criteria for
number of data points that were used in the fitting and some radius
for smoothing over a window of neighboring data density / number predicting out.
Args:
radius: List[int]
absolute: (bool) plot mean residuals by value or absolute value
exclude: (int) exclude model fits with under this many data points
"""
smoothed_residuals = self.get_smoothed_residuals(radius=radius)
smoothed_residuals = smoothed_residuals.loc[smoothed_residuals['num_data'] > exclude].copy()
for k, v in self.pv_groups.items():
plot_residuals(residual_array=v.residuals, group_name=k, absolute=absolute)
smooth = smoothed_residuals.loc[smoothed_residuals[self.col_group] == k]
smooth_mean = np.asarray(smooth[['far_out', 'num_data', 'residual_mean']])
smooth_std = np.asarray(smooth[['far_out', 'num_data', 'residual_std']])
plot_residuals(residual_array=smooth_mean,
group_name=f'{k} smooth mean radius {radius}', absolute=absolute)
plot_residuals(residual_array=smooth_std,
group_name=f'{k} smooth std radius {radius}', absolute=True)