Speaker:
Hanfeng Chen
Bowling Green State University

Title:
LARGE SAMPLE INTERVAL MAPPING METHOD FOR GENETIC
TRAIT LOCI IN FINITE REGRESSION MIXTURE MODELS           

Abstract:
Consider the large sample interval mapping method for genetic trait locus (GTL) under the finite mixture model with a non-linear regression kernel function. Identifiability problem of the model parameters along with consistency of their MLEs will be discussed in this talk. The asymptotic results on the likelihood ratio test for GTL detection will be presented. In particular, it will be shown that the threshold values or p-values required by the interval mapping method can be conveniently approximated via the null limiting distribution of the likelihood ratio test statistics. 

The large sample study presented in this talk has general applicability to the interval mapping method of GTL in genetic research. First, the organism populations derived from either a backcross or an intercross design are considered. Secondly, most commonly-used models in genetic research, such as exponential family mixture, logistic regression and generalized linear mixture models, are special cases of finite non-linear regression mixture models. Third, unlike all existing results with the finite mixture models, the boundness conditions on the parametric space are not imposed in the study. 

The talk is based on recent joint work with Drs. H. Zhang and Z. Li.