A second general feature of the power curves for Caucasian studies in Figure 2 is that aside from the Affymetrix 100 K chip (which is no longer available), there are not major differences in power across the other seven chips. For Caucasian samples the chips are typically ordered (with decreasing power): Illumina 1M, Illumina 650 k, Illumina 610 k, Affymetrix 6.0, Illumina 300 k, Affymetrix 500 k, but the absolute difference in power between the best and worst of these chips is often no more than around 10%. Put another way, for effect sizes in the range 1.3–1.5, a study with the Affymetrix 500 K chip would have the same power as one with the Illumina 1 M chip if its sample size were larger by 10–20%, with smaller increases in sample sizes giving studies with other chips the same power. Further, in Caucasian studies, power for all chips other than the Affymetrix 100 K chip is quite close to the best which could be obtained, namely by directly genotyping the causative SNP.
