Abstract
Twelve rice (Oryza sativa L.) varieties were cultivated over a four-year period from 1997 to 2000 to conduct a stability analysis of yield and yield characters, using the joint regression analysis and other stability measurements. From the regression analysis, all characters, with the exception of 100-grain weight, exhibited significant genotype x environment (G x E) interaction making it difficult to identify superior genotypes. The G x E interaction for grain yield was accounted for in part by both linear and non-linear environmental effects. The G x E interaction for plant height, maturity, and panicle length could be explained by additive linear environmental effects while that for tillering ability, effective tillering, grain weight per panicle, and spikelet fertility was due to non-linear effects. More genotypes were adjudged unstable, in respect of more characters, by the stability variance, ?2 relative to the regression coefficient, band the deviation mean square, S2di. From the estimates of the, ?2 and the S2di, IRAT 169, ITA 230, and SIPI 692033 were identified as unstable in yield and hence unsuitable for cultivation across seasons. However, ITA 230 and TOX 3226-53-2-2-2, which were considered unstable in yield by the stability variance, ?2i were selected for stability by YSi (the modified rank-sum statistic) in addition to TOX 3107, WITA 1, IR 8, and M 55. Seasonal differences in total rainfall up to the boot stage appeared to be the main environmental factor that influenced variation in genotypic performance.
