G-Methods for Time-Varying Data

Three main g-methods for time-varying treatments:

Method	Approach	Key Advantage
IP Weighting	Weight by treatment probabilities	Separates confounding model
G-Formula	Simulate counterfactual trajectories	Flexible outcome modeling
G-Estimation	Structural nested models	Handles effect modification

All give consistent estimates under:

• Correct model specification
• Sequential ignorability
• Positivity