A combination of statistical procedures was used to determine which yield components were most likely to be effective selection criteria for yield improvement in barley (Hordeum vulgare L.). The procedures included: 1) simple phenotypic correlations between grain yield and its endpoint components, 2) ridge regression to determine which characters were strongly interrelated, and 3) multivariate models and path analysis. Each procedure was applied in a 2‐year study of 25 near‐homozygous spring barley selections from ‘Coast’/‘Lion’//‘Winter Club’ and two check cultivars. Yield/plant (Y) was the primary measure of grain yield. Components measured were X1, spikes/plant; X2, kernels/spike; X3, kernel weight; and X4, kernel weight/spike. Two models were studied using original data and log transformed data: YA, involving X1, X2, and X3, and YB, involving X1 and X4. These relate to grain yleld/plant as YA = X1‐X2‐X3 and YB = X1‐X4 because X4 = X2X‐3. Phenotypic correlations gave unstable inter‐year and intercharacter relationships. X3 and X4 were strongly correlated with Y. Path analysis of model YA identified X3 and model YB identified X4 as having stronger direct effects on yield than other characters. X4, weight/spike, was judged as the primary character worthy of further investigation as a simply applied selection criterion. Results of the analyses with the log transformed data were completely consistent and practically identical with the analyses of untransformed data.

No abstract

No abstract

No abstract

No abstract