How does inverse probability weighting work?
IPW gives the most influence to treatment choices that were unlikely for a person’s measured baseline characteristics. The weighted treated and untreated groups can then represent the same target population.
Try it in the experiment
- Compare risk scores between groups before weighting.
- Select “Try IPW” to apply the fitted propensity-score weights.
- Compare balance and the weighted outcome difference with truth.
This experiment normalizes weights within each treatment arm. A useful IPW estimate still requires measured confounding control, positivity, consistency, and an adequate treatment model. Weighting cannot recover an unmeasured common cause.