Buehler-Martin Lecturer Series - March 31, 2005
University of Minnesota
School of Statistics
College of Liberal Arts
Established by Mr. and Mrs. Thomas Martin
in Memory of
Robert J. Buehler, Professor of Statistics (1963-1988)
Semiparametric
Methods for Gene-environment Case-control Studies When Gene and
Environment Are Independent in the Population.
Raymond J. Carroll
Department of Statistics
Texas A & M University
Thursday, March 31, 2005
3:30 PM, 115
Ford Hall
Minneapolis, East Bank Campus
Social at 3:00 PM, 300 Ford Hall
Abstract
We consider case-control studies of gene and environment interactions
using prospective logistic regression models. In a typical case-control
study, neither the intercept of the logistic regression nor the
population probability of disease can be identified. However, in many
cases it is reasonable to assume that genotype and environment are
independent in the population, possibly conditional on covariates to
account for population stratification. In such as case, we show that
the intercept and population probability of disease are identified. We
develop a semiparametric likelihood approach for this problem, showing
that it leads to more efficient estimates of gene-environment
interaction parameters than the standard approach. In addition, if the
probability of disease is known in the population, we show efficiency
gains for estimating gene-environment interactions, again in contrast
to the standard approach. Multiple extensions are discussed, including
to missing genotype data, haplotype data, and measurement error in
genotypes or environmental variables. Applications to two important
data sets are discussed.
This is joint work with Nilanjan Chatterjee (National Cancer
Institute).