causalis.scenarios.unconfoundedness.refutation.overlap.reliability_plot.plot_propensity_reliabilityplot_propensity_reliability
Plot a propensity calibration reliability diagram.
The reliability plot groups observations into propensity bins and compares predicted treatment probability with observed treatment frequency. In a well-calibrated propensity model, points should stay near the diagonal:
Parameters
- estimateCausalEstimate
Estimate with diagnostic data (
m_hat; optionallym_hat_raw,d).- dataCausalData, optional
Optional fallback source for treatment
dwhen not stored in diagnostic data.- n_binsint, default 10
Number of calibration bins used to build the reliability table.
- show_recalibrationbool, default True
Overlay logistic recalibration curve
sigmoid(alpha + beta * logit(p))when parameters are available.- annotate_metricsbool, default True
Annotate ECE and logistic recalibration parameters on the figure.
- axmatplotlib.axes.Axes, optional
Existing axes to plot on.
- figsizetuple, default (7.2, 6.2)
Figure size.
- dpiint, default 220
Dots per inch.
- font_scalefloat, default 1.10
Font scaling factor.
- point_colorcolor, optional
Marker color for binned reliability points.
- diagonal_colorcolor, default “0.35”
Color for the perfect calibration diagonal.
- recalibration_colorcolor, default “C1”
Color for the logistic recalibration curve.
- min_marker_sizefloat, default 35.0
Base marker area for non-empty bins.
- marker_size_scalefloat, default 250.0
Additional marker area scaled by bin count share.
- savestr, optional
Path to save the figure.
- save_dpiint, optional
DPI for saving.
- transparentbool, default False
Whether to save with transparency.
Returns
The generated figure.
Examples
Canonical target
causalis.scenarios.unconfoundedness.refutation.overlap.reliability_plot.plot_propensity_reliability
Sections