Causal graphs representing the 8 scenarios summarized in Appendix Table 4, showing the potential for bias in risk and causal effect estimates obtained by the ABMs when only the distribution of U differs between populations used for parameterization and/or inference. A is not a cause of Y, and U is not a cause of L or Y (A); A is not a cause of Y, and U is a cause of Y only (B); A is not a cause of Y, and U is a cause of L only (C); A is not a cause of Y, and U is a common cause of L and Y (D); A is a cause of Y, and U is not a cause of L or Y (E); A is a cause of Y, and U is a cause of Y only (F); A is a cause of Y, and U is a cause of L only (G); A is a cause of Y, and U is a common cause of L and Y (H).