Developing an estimating equation framework for robust summary data Mendelian randomization
PhD project (3/4 yr research project leading to independent research at the doctorate level)
Jack Bowden, Frank Windmeijer, 3rd Supervisor TBC
Mendelian randomization (MR) uses genetic data to probe questions of causality within the instrumental variable framework. In recent years it has become popular to perform Mendelian randomization studies using summary data estimates of gene-exposure and gene-outcome associations published by large international disease consortia
Methods originally derived from evidence synthesis such as inverse variance weighted (IVW) meta-analysis and MR-Egger regression have so far been used to estimate causal effects in the presence of invalid instruments. Novel robust methods such as the Weighted Median have also been employed. Whilst these approaches are superficially distinct, it is unclear how they are methodologically related, and indeed whether they are being implemented in an optimal fashion.
Aims & objectives
The aim of this PhD is to
1. Develop a unified estimating equation framework for summary data MR based on heterogeneity statistics that incorporates different model choices and loss function criteria.
2. Develop a general theory for assessing influential and outlying genetic variants.
3. Understand how previously proposed summary data methods represent special cases of this general framework, in order to propose fully efficient, novel implementations
The project will make use of instrumental variable methods as applied in MR, meta-analytical techniques employed in evidence synthesis and the generalized method of moments methodology from econometrics.
Bowden et al. Improving the accuracy of two-sample summary data Mendelian randomization: moving beyond the NOME assumption. bioRxiv. 2017
Windmeijer, F. Two-Stage Least Squares as Minimum Distance Working paper, department of economics, University of Bristol, 2017
Created on July 20, 2017, 12:52 p.m.