2.5 Data analyses

Analysis of variance was conducted on field data for two seasons using the restricted maximum likelihood (REML) approach implemented in the lme4 package in the R statistical environment. Genotypes and season were treated as random factors to estimate unbiased predictors from the model. Traits having non-significant genotypic effects were dropped, and the predicted genotype effects of different traits were used for further analyses. To identify the 10% extreme genotypes for overall phenotypic performance, trait predictors were ranked in descending order, individually for all traits, so that the highest value gets rank one. The cumulative ranking (rank sum) of each genotype was computed by adding the ranks of that genotype for all traits. For instance, under P+, the genotype ISe 1387 was ranked 1^st for PH, NL, LF, NC and TSE, 2^nd for LL, 3^rd for SPC and 37^th for NPT making its rank sum 47. The rank sums were similarly worked out for the rest of the genotypes. A similar approach was implemented for the greenhouse data.

From the rank sums, five extreme genotypes, having low- and high-rank sums, were picked under each category of P- and P+ treatments. Four classes of extreme genotypes were selected: lowest rank sum (high performers) under P+, highest rank sum (low performers) under P+, high and low performers under P-.