Nonparametric Estimation from Kin-cohort Data

• Activity Number: 166
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yuanjia Wang*+ and Daniel Rabinowitz
• Companies: Columbia University and Columbia University
• Keywords: kin-cohort ; missing covariats ; effciency bounds ; nonparametrics

Abstract

In kin-cohort studies, genotype and disease status are obtained from a group of volunteers (probands), and disease history is obtained from relatives of the probands. The relatives of the probands are not genotyped but form a retrospective cohort for the disease. Investigators are often interested in assessing the distribution of gene-related phenotypes as a function of deleterious mutation. For rare genetic variants, it can be difficult to obtain adequate number of carriers who develop disease for accurate estimation of the effect. Several parametric and semiparametric approaches have been proposed to estimate the influence of genotype on the distribution of disease status. Here, a class of nonparametric estimators is developed, and the optimal member of the class is derived. The optimal estimator in the class lies in the tangent space of the nonparametric model, and is therefore efficient among the class of regular unbiased estimators. A two-step approach to computing the estimator is described. Some simulation results are shown, and the method will be applied to Parkinson's disease.