Page 234 - Special Topic Session (STS) - Volume 4

STS582 Mariza de A.
perform the least-squares regression of the outcome Y on the predicted values
from the first regression. In the case that the IV is genetic information, the
predicted values are the means of X for each category. There are other
methods available such as limited maximum likelihood (LIML) (8) and
generalized method of momments (GMM) (9). The MR was also extended to
meta-analysis (10,11).

Figure 3: Mendelian Randomization Models

Approaches for Instrumental Variables (IV) analysis (1):
1. Ratio to coefficients method or the Wald method: It is the simplest way to
estimate the ausal effect of the exposure on Y (continuous or binary). IV
can be discrete, continuous and polytomous.
2. Two-stage least squares (2SLS) method: This consists of two regression
stages: the first –stage regression of the exposure on the genetic IVs; the
second-stage regression of the outcome on the fitted values of the
exposure from the first stage. It is known as Two-stage least squares (2SLS).
It can have multiple IVs. The outcome can be continuous and binary and
can use Generalized Linear Models (GLM).

223 | I S I W S C 2 0 1 9