Computer program for construction of regression function for phenotype in agroclimatic models with interactions

Abstract

Regression models that connect agronomic traits to climatic factors provide valuable insights into phenological characteristics of cultivars. The genotype-by-environment interactions are modeled by a weighted sum of pairwise products between a control functions and group indicator variables. In contrast to existing modeling frameworks in our approach the analytic form of a control function, regression coefficients and a set of predictors are inferred by stochastic minimization of the deviation of the model output from data. The approach was successfully applied to the three datasets for soybean and chickpea to predict time to flowering with coefficient of determination 0.45-0.97.