In the current study, we compare propensity score (PS) matching methods for data with a cross-classified structure, where each individual is clustered within more than one group, but the groups are not hierarchically organized. Through a Monte Carlo simulation study, we compared sequential cluster matching, preferential within cluster matching, greedy matching, and optimal full matching, using propensity scores from four different models. We manipulated the correlation of random effects between crossed factors, degree of cross-classification, intra-unit correlation coefficient, variance explained by covariates, level-1 and level-2 sample sizes, and proportion treated. The results indicated that the four matching methods performed similarly when PSs were estimated with logistic regression containing both level-1 and level-2 covariates. Also, the matching methods were robust to the omission of all level-2 covariates if the PS model was a logistic cross-classified random effects model, which included random effects at both clustering levels.

In the current study, we compare propensity score (PS) matching methods for data with a cross-classified structure, where each individual is clustered within more than one group, but the groups are not hierarchically organized. Through a Monte Carlo simulation study, we compared sequential cluster matching, preferential within cluster matching, greedy matching, and optimal full matching, using propensity scores from four different models. We manipulated the correlation of random effects between crossed factors, degree of cross-classification, intra-unit correlation coefficient, variance explained by covariates, level-1 and level-2 sample sizes, and proportion treated. The results indicated that the four matching methods performed similarly when PSs were estimated with logistic regression containing both level-1 and level-2 covariates. Also, the matching methods were robust to the omission of all level-2 covariates if the PS model was a logistic cross-classified random effects model, which included random effects at both clustering levels.