A Random-Effects Regression Model Based on the Odd Log-Logistic Skew Normal Distribution

Regression models with random effects are quite common when the data have a longitudinal or grouped structure for more flexible estimation of the correlation between or within individuals. The normality assumption may not be realistic and can hide important characteristics of variation between and within individuals. We consider that the response variable follows the new odd log-logistic skew-normal distribution, which includes as special cases the normal and skew-normal. The estimates of the model parameters are calculated by the maximum likelihood method. We illustrate the usefulness of the proposed random-effects regression by means of a real longitudinal data set on the postharvest storage of lychee. We evaluate the physicochemical, physiological, biochemical and sensorial aspects associated with the techniques of postharvest lychee conservation.