A tutorial on statistical methods for population association studies

Top Cited Papers
Publisher Website
Google Scholar
Abstract
Although population association studies are not new, there remain many areas of disagreement over appropriate statistical analyses. This article provides an overview of statistical methods, including areas of controversy and ongoing developments. It does not consider family-based association studies, nor linkage or admixture studies. I first cover analyses that are preliminary to association testing: testing for Hardy–Weinberg equilibrium; imputing missing genotype data; inferring haplotype from genotype data; measures of linkage disequilibrium and estimates of recombination rates; and choosing tag SNPs. Among tests of association, I cover case–control, quantitative and ordered phenotypes, and analyses that are based on single SNPs, multiple SNPs and haplotypes. There is a discussion of issues that are relevant to genome-wide association studies. I discuss Genomic Control and other approaches to the problem of population stratification. I give particular attention to the problem of multiple testing, and discuss both frequentist and Bayesian approaches to addressing the problem.