Issues in Diallel Analysis

Abstract

Critical issues in the use of diallel analysis are reviewed. From a statistical point of view the critical issue concerns the choice of a model with fixed or random genotypic effects. From a genetical point of view, two assumptions are critical in attempts to interpret the resuits of diallel analyses. The assumption concerning the independent distribution of genees in the parents is most critical to proper interpretation and seems to be least acceptable in actual practice. The second assumption, that there is no epistasis, may frequently be incorrect. Epistasis affects estimates of general and specific combining ability mean squares, variances, and effects in an unpredictable manner. As an alternative to genetic interpretation, the statistical description provided by dialiel analysis can be used to help answer questions concerning the importance of specific combining ability and the predictability of hybrid performance using general combining ability or parental performance.