bels: backward elimination locus selection for studies of mixture composition or individual assignment

Methods of evaluating loci in studies of mixture composition or individual assignment are largely based on performance characteristics of individual loci. Synergisms between loci are not exploited. Loci are often evaluated based on their ability to resolve individual populations, even though multipopulation aggregations are more commonly of interest. In addition, measures of locus performance may indirectly relate to investigative objectives. A new computer program, bels, offers an alternative that addresses these limitations and may be preferable to existing methods in some applications. The algorithm is illustrated using Yukon River chum salmon (Oncorhynchus keta) data.