Additive main effects and multiplicative interactions analysis of harvest index performances in cassava (<i>Manihot esculenta</i>, Crantz) genotypes across 4 environments

Abstract

Eight cassava genotypes were evaluated for harvest index performance across four environments. Data analysis was performed using MATMODEL and GGEbiplot. AMMI analysis of variance showed that 10.02% of the total sum of squares was attributable to environmental effects, 3.99% to genotypic effects and 50.13% to GEI effects. The GEI sum of squares contained approximately 76.52% (0.30709) pattern and 23.48% (0.09442) noise of the total GEI. The mean squares for IPCA 1 and IPCA 2 were significant at P = 0.000 and 0.002 respectively; all together they contributed 94.18% of the total GEI. Therefore, the post-dictive evaluation using an F-test at 0.000 and 0.002 suggested that two principal axes of the interaction were significant for the model with 16 degrees of freedom. The predictive assessment measured by the average root mean square predictive difference (RMS PD), selected AMMI1 with the first interaction PCA axis as the most predictively accurate. The AMMI1 model had the lowest average RMS PD (9.996). Mean performance and stability of the genotypes assessed by biplot analysis showed that the most stable genotypes were G2, G1 and G7. However G5 was highly unstable followed by G8 and G3. Two mega-environments were defined namely: G5-winning niche and G3-winning niche. The current study has demonstrated that the GGE biplot is a useful tool for the analysis of multi-environment trial (MET) data.

Keywords: AMMI, GGEbiplot, stability, harvest index, mega-environment