Inverse probability weighting

Reweight treated and untreated people using their fitted chances of receiving the treatment they actually received.

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

  1. Compare risk scores between groups before weighting.
  2. Select “Try IPW” to apply the fitted propensity-score weights.
  3. 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.