Research and Evaluation Exam 1 Practice

Propensity score matching aims to create balance in observed covariates between treated and untreated groups in observational data by pairing individuals who have similar probabilities of receiving the treatment. The propensity score is the estimated probability of treatment given the covariates, often computed with a logistic regression. By matching on this score, the distribution of the observed covariates becomes similar across groups, so differences in outcomes are more plausibly attributed to the treatment rather than to confounding factors. After matching, you compare outcomes within the matched pairs or sets, which helps approximate what randomization would achieve in an experiment. Keep in mind this only balances observed covariates included in the model; unmeasured confounding can still bias results. Estimation typically involves calculating propensity scores and then matching (nearest neighbor, with or without a caliper), followed by outcome analysis on the matched data. The other descriptions don’t capture this idea: weighting changes the sample via weights, randomizing after measuring outcomes isn’t how analyses of observational data work, and a simple test that ignores covariates misses the adjustment for confounding.