Best Linear Unbiased Prediction (BLUP) Nilai Pemuliaan Sifat-Sifat Kuantitatif Kacang Hijau (Vigna radiata L. Wilczek) Berbasis Sebaran Peluang

Abstract

Mung bean is the third most important food legume in Indonesia after soybean and groundnut. Accurate prediction of breeding values remains a major challenge in mung bean breeding, particularly for quantitative traits influenced by genotype–environment interactions. This study aimed to predict breeding values using the Best Linear Unbiased Prediction (BLUP) approach based on data probability distributions. Seven local lines and five national superior varieties were evaluated using an incomplete split-plot randomized design. The assumptions of normality and homogeneity of variance were tested using the Anderson–Darling and Levene’s tests on the residuals of the normal distribution model. When assumption violations occurred, the data were subsequently analyzed using a restricted mixed linear model with a log-link function across several exponential family distributions: normal, gamma, geometric, and exponential. The best-fitting distribution was determined based on the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and corrected AIC (AICc) values. The results indicated that the assumptions were violated, and the gamma distribution provided the best BLUP estimation for non-normal and heterogeneous data. These findings suggest that incorporating probability distribution approaches enhances the validity and efficiency of BLUP-based breeding value prediction in mung bean improvement programs