Unit gamma mixed regression models for continuous bounded data. Issue 6 (13th April 2023)

• Record Type: Journal Article
• Authors: Petterle, Ricardo R.; Taconeli, César A.; da Silva, José L. P.; da Silva, Guilherme P.; Laureano, Henrique A.; Bonat, Wagner H.
• Is Part Of: Journal of statistical computation and simulation. Volume 93:Issue 6(2023)
• Page Start: 1011
• Page End: 1029
• Publication Date: 2023-04-13
• Subjects: Body fat percentage -- random effects -- automatic differentiation -- laplace approximation -- water quality index -- template model builder; Linear inference -- regression; Mathematical statistics -- Data processing -- Periodicals; Digital computer simulation -- Periodicals; 519.5028505
• DOI: 10.1080/00949655.2021.1970164
• Languages: English
• ISSNs: 0094-9655

Abstract

Abstract : We propose the unit gamma mixed regression model to deal with continuous bounded variables in the context of repeated measures and clustered data. The proposed model is based on the class of generalized linear mixed models and parameter estimates are obtained based on the maximum likelihood method. The computational implementation combines automatic differentiation and the Laplace approximation (via Template Model Builder/C++) to compute the derivatives of the log-likelihood function with respect to fixed and random effects parameters. We carry out extensive simulations to check the computational implementation and to verify the properties of the maximum likelihood estimators. Our results suggest that the proposed maximum likelihood approach provides unbiased and consistent estimators for all model parameters. The proposed model was motivated by two data sets. The first concerns the body fat percentage, where the goal was to investigate the effect of covariates which were taken in the same subject. The second data set refers to a water quality index data, where the main interest was to evaluate the effect of dams on the water quality measured on power plant reservoirs. The data sets and R code are provided as supplementary material.