Detection and estimation of genotype-environmental, linkage and epistatic components of variation for a metrical trait

Abstract

1. Biometrical genetical models, scaling tests and sequential model fitting procedures are described for detecting, classifying and estimating the contributions of genotype-environmental interactions, linkage and epistasis when a simple model, which assumes their absence, fails to account for all the observed variation within the generations which can be derived from an initial cross between a pair of inbred strains. 2. The use of these procedures when all three causes of failure of the simple model are present simultaneously is illustrated by the analysis of 36 variances, for each of two characters, obtained from 21 generations of a cross between varieties 1 and 5 of NIcotiana rustica. 3. The results of the analyses show that the scaling tests are in general more sensitive than sequential model fitting procedures for detecting and classifying linkage, genotype-environmental interactions and epistasis when all models fail, although estimates of their contributions to the variances can only come from model fitting. 4. Comparison of these results with those of earlier investigations of the same cross leave no doubt of the greater sensitivity of both the analyses and the experimental design used here. 5. The advantages and disadvantages of biometrical genetical analyses of generation means and within generation variances are discussed and illustrated from the present analyses and an earlier analysis of the means of the same experiment. It is stressed that the two are complementary and not alternatives and extrapolation from the control of a character revealed at one level to that at the other can be misleading. © 1970 The Genetical Society of Great Britain.