Approximate likelihood inference for haplotype risks in case-control studies of a rare-disease.

Zhijian Chen successfully defended his thesis entitled "Approximate likelihood inference for haplotype risks in case-control studies of a rare-disease" on 16 August 2006.
 

The standard study design to study risk factors for rare diseases is the case-control design. Genetic association case-control studies often include haplotypes as risk factors. Haplotypes are not always observed, which leads to analysis of data with missing covariates. Maximum likelihood (ML) inference is then based on the solution to a set of weighted score equations. However, the weights can not be calculated exactly. We describe three methods that approximate ML by approximating the weights: i) naive application of prospective ML (PML), which ignores the case-control sampling design, ii) an estimating equations (EE) approach and iii) a hybrid approach which is based on PML, but with improved weights suggested by the EE approach. We investigate the statistical properties of the three methods by simulation. In our simulations the hybrid approach gave more accurate estimates of statistical interactions than PML and more accurate standard errors than the estimating equation approach.

This type of interdisciplinary work is a hallmark of our program in Applied Statistics at Simon Fraser University. For more information, please contact Zhijian Chen (zhijianc@sfu.ca) or his supervisor Brad McNeney (mcneney@stat.sfu.ca), Department of Statistics and Actuarial Science.

30 August 2006.