Model selection in additive main effect and multiplicative interaction model in durum wheat

The study included data set of 20 durum wheat genotype across 15 rain-fed environments. A combined analysis of variance showed that the genotypes differed significantly for seed yield and GE (year × location) interaction. Cross validations procedure and four various F-tests including FGollob, FRatio, FGH1 and FGH2 are used for testing the GE interaction principal component analysis (IPCA) axes and indicated that two, four, six or seven axes could be significant. According to EV1, D1, AMGE1 and SIPC1 parameters, genotypes G3, G7 and G17 were the most stable genotypes while based on EV4, D4, SIPC4 and AMGE4 parameters, genotype G13 was the most stable genotype. The hierarchical clustering showed that the twenty one studied the AMMI stability parameters and mean yield could be divided into four distinct groups. Group III contains mean yield, SIPC4, SIPC6 and SIPC8 which were computed from four, six or eight IPCAs. In conclusion, G13 (DON-MD 81- 36) was found to be the most stable genotype as well as high mean yield performance (2592.45 kg ha-1) and so is recommended for commercial release in semi-arid areas of Iran. Also, the SIPC-based stability parameters of the AMMI model was found to be useful in detecting the yield stability of the genotypes studied.