• Propensity score methods provide a scalar balancing score that, when used for matching, stratification, or weighting, aim to remove bias from all observed covariates in observational causal analyses [1].

• Inverse probability weighting and semiparametric regression enable estimation of causal effects in longitudinal data with missingness or time-varying confounding, without fully specifying the full likelihood [4], [20].

• Diagnostics and validation for survival and time-to-event causal analyses rely on martingale-based residuals, goodness-of-fit assessments, and residual plots to check proportional hazards and model adequacy [5], [11], [15].

• Sensitivity analyses address robustness to unobserved confounding and model misspecification, illustrating bounds and scenarios under which causal conclusions may change [16], [14], [6].