A phenotypic stability analysis of common bean (Phaseolus vulgaris L.) grain yield (kg.ha-1) was carried out using data of the Regional Trial of the State of Paraná during the year/harvest 1996/97. A non-linear model using the parameters proposed by Toler, Yij = alphai + [Zjbeta1i + (1 - Zj) beta2i] muj + epsilonij, was adopted. The author proposed tests of the hypotheses H(beta1i = beta2i) and H(beta1i = beta2i = betai = 1) according to the principles of regression analysis, that allow the genotypes to be placed in five groups (A, B, C, D, and E) based on their response pattern. The parameters of the model are simultaneously estimated using iterative (non-linear) least squares through the modified Gauss-Newton method. The model was useful in classifying the genetic materials according to their yield and response patterns, contributing to a greater understanding of their yield behavior. The mean of the A and E groups was influenced by a negative association between <img src="http:/img/fbpe/brag/v59n1/a15b1.gif" alt="a15b1.gif (192 bytes)" align="absmiddle"> and <img src="http:/img/fbpe/brag/v59n1/a15b2.gif" alt="a15b2.gif (190 bytes)" align="absmiddle">. The majority of genotypes with a double favorable response pattern (A group, convex pattern) had lower yields. Genotypes with higher yields were placed mostly in the E group (concave pattern, double unfavorable). Those from groups with a single-segment pattern (B, C, and D) showed variable yields. The difficulty in finding genotypes with both high yields and convex response pattern was evident.
Phenotypic stability; response patterns; bean
