I calculated the IHS statistics (the method is desribed in "A Map of Recent Positive Selection in the Human Genome". DOI: 10.1371/journal.pbio.0040072). This statistics measures ongoing positive selection signals, and in the paper they use the proportion of high IHS sites (i.e., the absolute value of IHS higher than 2) to evaluate the strength of ongoing positive selection signals in a genome window or gene. Since it measures on going positive selection signals, the rare and fixed SNPs are ignored, and only calculate IHS for common variances (those whose minor allele frequency is higher than 0.05).
Here my question: I use 1000 genome data. For a gene that I am interested, if I calculate the IHS statistics on the whole population, I found among the 400 SNPs in the gene more than half of them show strong selection signal (|IHS|>2). The result indicates the gene is under positive selection among the whole population. But when I calculate such IHS in each sub populations (E.g., African, American), I did not found positive selection signals in the gene for each of sub population (only a few IHS higher than 2 in each sub population, which is close to random levels by permutation test). The results confused me, because after I found the gene is under selection in the whole population, I want to see which population shows the strongest selection signal. But I did not found selection signals in each sub population. I am not sure how to explain this.
Any ideas or suggestions about this?