Propensity Score Matching

A propensity score (傾向分數) is each unit's estimated probability of receiving the treatment given its measured characteristics. Matching treated and untreated units that share a similar propensity score builds comparison groups that look alike on those characteristics, mimicking a randomized experiment.

It is a convenient way to adjust for many confounders at once by collapsing them into a single number. The crucial caveat: it can only balance confounders you measured and included. It does nothing about unmeasured confounding, so a tidy propensity-score analysis still rests on the strong (and untestable) assumption that you have captured all the important confounders.