View article

Comparing the amount of geographic variation in allele frequencies for protein and DNA polymorphisms is a powerful method for detecting the effects of selection. Some statistical artifacts must be kept in mind however. Simulations indicate that estimators of Wright’s F _ST are much better measures of geographic variation than are genetic distance measures; that pooling alleles so that all polymorphisms are treated as two-allele polymorphisms is sometimes necessary to avoid statistical artifacts; and that for a given total sample size two or three population samples can be just as efficient at detecting selection as a larger number of smaller samples.