A1146
Title: Beta four parameter generalized linear mixed model using a Bayesian approach to predict paddy productivity
Authors: Dian Kusumaningrum Hermanto - Prasetiya Mulya University/IPB University (Indonesia) [presenting]
Abstract: A new Generalized Linear Mixed Model (GLMM) for a response variable having a beta four-parameter distribution based on the Bayesian approach is introduced. The framework expanded the beta four-parameter regression model to incorporate random effects in the model. The methodology is illustrated through simulations and applied to predict paddy productivity. Paddy productivity ranges between a minimum and maximum value; therefore, assuming a beta four-parameter distribution is most appropriate. Farmer survey and Sentinel satellite imagery data were used as covariates. The response variable was based on plots surveyed by the Central Bureau of Statistics (CBS) in Central Kalimantan, Indonesia. Results showed that this approach could overcome complicated back-transform processes, difficulties in interpreting the results obtained, and bias in the estimated parameters if the transformation to a standard beta distribution process was to be applied. In predicting paddy productivity, results were proven to be more accurate. Thus, it will be a beneficiary as an early warning system for food insecurity, a reference for the food self-sufficiency program, and a basis to calculate premiums and risks for the alternative Area Yield Index (AYI) crop insurance policy.
