Mendelian randomization in the presence of pleiotropy

Automne 2016
Mendelian randomization
in the presence of pleiotropy
Séminaire de statistique
Conférencière: Celia Greenwood
Date, heure et endroit
Senior Investigator, Lady Davis Institute
Associate Professor, McGill University
Jeudi 24 novembre 2016
VCH-3830
13h30
Résumé
Mendelian randomization is a popular type of instrumental variable analysis often
used to make causal inference in the field of genetic epidemiology. For example,
a genetic variant might be used as an instrument in order to infer whether a
metabolic measure or a gene expression measure is causally related to disease risk.
However, genetic variants will be invalid instruments when they act on more than
one intermediate measure, and this is very common. I will describe Mendelian
randomization, and show a simple example where it has worked well. Then, when there
are multiple genetic variants that could be used as instrumental variables, I will
describe approaches to reducing bias in the presence of pleiotropy. Simulations
compare existing methods and a new method we call “constrained instrumental
variables (CIV)
Ting-Huei Chen & Louis-Paul Rivest
Responsables du séminaire de statistique