Understanding Deeply · 01
IPTW vs propensity score matching.
Start the guided tour, or open “Follow the arithmetic” to trace a patient's outcome into the group averages and final estimate.
From patient differences to a treatment comparison
Real-world evidence often involves comparing groups that differ before treatment begins. This fictional example shows how those differences can affect a raw comparison of recovery scores.
Propensity score matching selects comparable pairs. Inverse probability of treatment weighting (IPTW) changes each observation’s contribution to the group average. The recorded outcomes stay the same; the patients included or their influence on the estimate change. Explore balance, overlap and the arithmetic alongside the result.
These methods address measured differences under assumptions explained in the playground. They do not automatically remove unmeasured confounding. For another part of the research question, explore intercurrent events and estimand strategies.
The simulator uses invented patients and a known fictional treatment effect. Its method notes explain the calculations, target populations and assumptions.
Next: intercurrent events ↗