Grain yield is complex traits, influenced by genetics and the environment, posing challenges for prediction. The goal of the present study was to identify key traits that contribute to rice yield by applying seven statistical techniques. The study examined twenty-four rice genotypes following a randomized complete block design (RCBD) with three replications. Pearson's correlation analysis revealed several traits exhibiting significant positive correlations with grain yield (r≥0.60), including thousand-seed weight (r = 0.63), filled seeds/panicle (r = 0.74), and number of panicles/m2 (r = 0.70). Multiple linear regression identified significant predictors as the number of panicles/m2 (R=0.01) and thousand-seed weight (Coefficient=0.12). Stepwise linear regression suggested key yield indicators viz. the number of panicles/m2 (R=0.01), filled seeds/panicle (R=0.02), thousand-seed weight (R=0.11), and panicle length (R=0.13). Bayesian linear regression found similar traits with a Bayes factor of 995.2 and an R-squared value of 0.77. Exploratory factor analysis showed grain number/panicle, filled seeds percentage, filled seeds/panicle, thousand-seed weight, grain width, and panicles/m2 are highly influenced grain yield and explained 61.9% total variance. Principal components analysis revealed the first three components explaining 79.1% of yield variation, with thousand-seed weight, grain width, and panicle number/m2. Path analysis demonstrated the number of panicles/m2, filled seeds/panicle, thousand-seed weight, and panicle length have large and significant positive direct effects on grain yield. These findings strongly suggest that selecting breeding materials with traits like high panicle density/m2, larger panicle size, more filled seeds/panicle, and higher thousand-seed weight, can significantly increase the rice yield.
Key words: Rice yield components, Stepwise regression, Bayesian linear regression, Exploratory factor analysis, Path analysis.
|
