Semiparametric Inference of Untyped Variants on Right-Censored Outcomes

• Activity Number: 604
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Zhiguo Li*+ and Yu Jiang and Janice McCarthy and Andrew Allen and Kouros Owzar
• Companies: Duke University and Duke University and Duke University and Duke University and Duke University
• Keywords: censored outcome ; genome-wide association study ; proportional hazards model ; untyped SNPs

Abstract

We propose semi-parametric methodology for analyzing the effect of untyped variants on right-censored phenotypes (e.g., time to death). Assuming a proportional hazards model for the variants and environmental factors, we use an estimating equation where the contribution of each subject is the average of the contributions in a partial likelihood score equation for all possible variants, weighted by the probabilities of the variants given the known genotype. In contrast to the two-step approach of plugging the imputed variants, as if they were actually observed, into the proportional hazards model, our proposed method properly accounts for the variability of the imputation process. In contrast to the approach of arbitrarily dichotomizing the outcome at a landmark, so as to use methods for inference of untyped variants with binary outcomes, our approach properly accounts for censoring. The asymptotic properties of the resulting estimator are established. Simulation studies will be conducted to assess the finite sample performance. We will illustrate the proposed methodology by conducting a genome-wide analysis to identify variants associated with overa