Chapter 4:

In this chapter, we discuss estimation of causal effects using instrumental variables for both continuous and binary outcomes based on individual-level data. We present the ratio and two-stage methods, which assume that genetic variants are all valid instrumental variables. While there is a dazzling array of more sophisticated methods for IV analysis, the ratio and two-stage methods are foundational to the understanding of Mendelian randomization. We provide code for implementing the methods using standard statistical software packages.