that are carefully prepared from pools of DNA from 20 individuals. Quality control evaluations come chiefly from assessment of the array-to-array variation in hybridization intensities noted for replicate experiments, as well as assessments of pool-to-pool variation, as noted here. These assessments thus do not measure the features that are assessed by Affymetrix or Illumina software packages, but rely on estimates of variability in relation to the signals obtained. The 0.98+ correlation between observed and expected allele frequencies provides a modest difference from a set of perfectly accurate individual genotypes. 3a) The sample sizes in studies currently listed in dbGAP are larger, with an average size of 2,155 cases and controls 3b) The sample size here of 1,620 is divided into two separate case vs control comparisons, providing, as we note, moderate power to detect replicable modest-sized effects and lower power to detect very small effects. True effects that provide nominal statistical significance in clustered SNPs in only one sample represent false-negative findings in our first analysis. 4a) Many of the results of individually-genotyped studies represented in dbGAP are analyzed based on the assumption that the detailed haplotype structures identified in data from CEPH and Yoruban individuals will provide accurate representations
