Robust estimation of a marginal causal effect on the binary outcome using propensity score matching

초록

Observational studies encounter self-selection bias due to non-randomized treatment assignments. To solve the problem of the self-selection bias, we employ the propensity score matching approach to compare non-randomized treatment and control groups and estimate the robust marginal treatment effect. In this study, we consider the double-adjustment and g-computation methods to reduce residual confounding bias, target bias, and propensity score model bias, and the caliper method to address the confounding bias. We compute the robust marginal causal effect of smoking on depressive symptoms to investigate the relationship between smoking and depressive symptoms using the 8-th Korean National Health and Nutrition Examination Survey data.

키워드