Using the Polynomial Power Method to Simulate Data with Correlated Non-Normal Error Structure for Assessing Methods Proposed for Family-Containing Genetic Association Tests

• Activity Number: 523
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): T. Beasley*+ and Russell Wolfinger and Wendy Czika and Hemant K Tiwari and David B. Allison
• Companies: The University of Alabama at Birmingham and SAS Institute and SAS Institute and The University of Alabama at Birmingham and The University of Alabama at Birmingham
• Keywords: Mixed Models ; Pedigree Data ; Simulation

Abstract

Human studies that examine the association between genetic polymorphisms and quantitative traits often use a general linear model approach which assumes independence of residuals. In contrast, samples which include related individuals requires a testing method which does not make this assumption for valid inference, and a number of seemingly appropriate methods have been presented in the literature, including various linear mixed model approaches. However, there is no consensus on which approach is preferable in terms of power, robustness, and computational ease, especially in the context of genome-wide association studies or departures from the normality assumption. To permit comparisons among these methods via computer simulations requires generating large-scale complex datasets with correlated residuals, itself a challenging task. To efficiently simulate such datasets, we propose a Polynomial Power Method to Generate Correlated Non-Normal Error Distribution for pedigree data. We describe SAS/IML code that generates genotypes via the "gene-dropping" method and data (outcome, covariate) for n individuals nested within K families with potential measurement error and missingness.