Causal inference with G-computation
Vincent Arel‐Bundock · 2025
As we will see below, G-computation estimates are equivalent to the counterfactual predictions and comparisons discussed in Chapters 5 and 6 . They are also closely related to the estimates we can obtain by Inverse Probability Weighting (IPW), another popular causal inference tool. IPW and G-computation impose similar identification assumptions, with one key difference: the former models the process that determines who gets treated, whereas the latter models the outcome variable.2 The next section introduces some key estimands that analysts can target via Gcomputation: the average treatment effect (ATE), average treatment effect on the treated (ATT), and average treatment effect on the untreated (ATU). Subsequent sections present the estimation procedure as a sequence of three steps: model, impute, and compare.