Adjusting for Population Heterogeneity and Mis-specified Haplotype Frequencies when Testing Nonparametric Null Hypotheses in Statistical Genetics

• Activity Number: 84
• Type: Invited
• Date/Time: Monday, August 12, 2002 : 10:30 AM to 12:20 PM
• Sponsor: JASA, Theory and Methods
• Author(s): Daniel Rabinowitz*+
• Affiliation(s): Columbia University
• Keywords: genetic epidemiology ; complete minimal sufficient statistic ; haplotype ; association studies ; linkage analysis ; missing data

Abstract

Population heterogeneity is a potential source of confounding in genetic epidemiology. Unlike observational studies generally, in genetic epidemiology, exact methods may be derived for adjusting for confounding due to heterogeneity. These exact methods are founded on the randomness inherent in the transmission of genetic material from parents to children, and, in the simplest cases, involve conditioning on parental genotypes. This conditioning strategy is not available, of course, when parental genotypes are missing or incomplete. However, a basis for extending the strategy to incomplete data lies in noting that parental genotypes are complete minimal sufficient statistics for the nuisance parameters; in many cases, non-trivial minimal sufficient statistics can be computed and conditioned to avoid confounding. Unfortunately, in many situations with missing or incomplete data, the minimal sufficient statistics are not complete, and conditioning strategies may fail to exploit available information. In such cases, score statistics that make use of all of the available information may be derived as the result of regression calculations.