Robust Inference in Mendelian Randomization with Application to Social Determinants of Alzheimer's Disease
Hyunseung Kang, Ph.D.
University of Wisconsin-Madison
Mendelian randomization (MR) is a popular method in epidemiology to estimate the causal effect of an exposure on an outcome using genetic variants as instrumental variables (IV). Often, two-sample summary data is used in MR where in one sample, summary statistics about the marginal correlations between the IVs and the exposure are available and in another sample, summary statistics about the marginal correlations between the IVs and the outcome are available. Unfortunately, many methods in MR are biased under weak or invalid instruments, where the correlation between the IVs and exposure is small or if the instruments have a a direct effect on the outcome. In this work, we study and propose a set of estimators and tests for the exposure effect that are robust to weak or invalid instruments. In particular, we show that the current popular method in the literature is not robust to weak instruments and propose several methods to fix the bias arising from weak instruments. Additionally, we present a novel estimator that can handle both weak and invalid instruments based. We conclude by applying our methods to study the effect of education on risk factors for AD.
Thursday, March 12, 2020
3:30 p.m. - 4:45 p.m.
Helen Wood Hall - Room 1W-501