statsmodels.treatment.treatment_effects.TreatmentEffect.overlap_summary#

TreatmentEffect.overlap_summary()[source]#

Summarize unclipped propensity scores by treatment group.

Returns:

DataFrame: Rows control and treated contain the observation count (nobs), minimum, 25th percentile, median, 75th percentile, maximum, and counts strictly below and above ps_bounds (n_below and n_above).

Notes

Uses predictions from the selection model before clipping. Extreme scores can indicate poor practical overlap; these summaries do not establish the population overlap assumption. No observations are dropped and no model is refitted.