Institute of Agricultural Research Statistics, Library Avenue, New Delhi-110012
*Present address: Central Rice Research Institute, Cuttack-6.
Online published on 25 January, 2012.
In complete diallel cross (CDC), with n inbred lines, there would be n(n-1)/2 crosses, excluding selfings and reciprocal crosses. As n increases, this number increases rapidly. With limited resources, it may not always be possible to handle all the crosses. In such cases one can perform only a sample of all possible crosses among the inbred lines. Such diallel crosses are known as partial diallel crosses (PDC). The method of analysing CDC and PDC with several characters have been applied to data on wheat (7 × 7 diallel) and triticale (8 × 8 diallel) to estimate heritabilities, genetic correlations and efficiencies of discriminant function. In the case of wheat, 14 crosses out of 21 were sampled according to design of Kempthorne and Curnow (K and C) with s=4, k=2. In the case of Triticale, four different sampling designs viz. factorial design (n=8, s=4) group divisible design (n=8, s=6), K and C design with n=8, s=5, k=2 and K and C design with n=8, s=3 and k=2 were used.
In the case of wheat data, standard errors of heritability estimates for characters like plant height, 1000 grain weight and number of grains per spike actually decreased with the use of PDC. The trend in the standard errors of genetic correlations due to the use of PDG was rather erratic. The PDC was found to have a higher percentage efficiency of selection based on discriminant function over straight selection.
The Triticale data, being available over two years, afforded additional information on the differences in the estimates of genetic parameters between two years. Existence of interaction between additive effects of genes and environmental differences over the years were found. In regard to the reduction in the standard error of the estimate of heritability, group divisible design appeared to be better than the other designs. It was further found that the reduction in the standard error of the genetic correlation due to the use of PDC was possible with all the four designs provided characters were yield per plant and days to flower. The factorial design gave the maximum percentage reduction in this case. The selection based on discriminant function involving yield per plant, days to flower, plant height and number of grains per ear was found to give higher expected genetic advance in both the years and in all the four designs. In regard to the percentage efficiency over straight-selection, factorial design appeared to be the best.
