Fig. 2From: Bias amplification in the g-computation algorithm for time-varying treatments: a case study of industry payments and prescription of opioid productsCausal diagrams for simulation studies. T1: treatment at time 1; T2: treatment at time 2; X1: Common cause of treatment at time 1, treatment at time 2, and outcome; X2: Time-varying confounder (i.e., confounder between treatment at time 2 and outcome affected by treatment at time 1); X3: Common cause of treatment at time 1 and treatment at time 2 (X3As, IV in scenario A; and X3B, near-IV in scenario B)Back to article page