Mendelian Randomization, 2nd Edition

4

Estimating a causal effect from individual-level data

In this chapter, we discuss estimation of causal effects using instrumental variables for both continuous and binary outcomes based on individual-level data. We focus attention on the case of a single continuous exposure variable, as this is the usual situation in Mendelian randomization. The same methods can be used in the case of a single binary exposure, although there are some nuances regarding the interpretation of the estimate (Section 8.11). We here present the ratio and two-stage methods, which assume that genetic variants are all valid instrumental variables. More sophisticated methods relaxing this assumption are introduced in later chapters. We also provide code for implementing ...