The potential p-value bias test.

<p>Consider the Fisher exact test p-value associated with an OR estimate for a hypothetical genotype present nowhere other than on every subject within a given spatial sphere. Each table entry reports the smallest such p-value, over all the subject-centered spatial spheres with the stated weight. These p-values provide a comparative measure of the potential for p-value bias due to population structure, before and after cohort homogenization. The value of the potential p-value bias test is the smallest of these p-values across all the tested sphere weights (rightmost column). Values shown pertain to application of the test to the original and homogenized (under the <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0048653#pone-0048653-g004" target="_blank">Figure 4</a> selected parameter values) Hamza et al. cohort. Note: for the homogenized cohort, the Fisher test is performed on the contingency table entries rounded to the nearest integer.