Likelihood-based inference for genetic correlation coefficients

We review Wright's original definitions of the genetic correlation coefficients F-ST, F-IT, and F-IS, pointing out ambiguities and the difficulties that these have generated. We also briefly survey some subsequent approaches to defining and estimating the coefficients. We then propose a general framework in which the coefficients are defined, their properties established, and likelihood-based inference implemented. Likelihood methods of inference are proposed both for bi-allelic and multi-allelic loci, within a hierarchical model which allows sharing of information both across subpopulations and across loci, but without assuming constancy in either case. This framework can be used, for example, to detect environment-related diversifying selection. (C) 2003 Elsevier Science (USA). All rights reserved.