Identifying change points for linear mixed models: a solution through evolutionary algorithms

Abstract

Linear Mixed Models; Change points; Evolutionary Algorithms. ABSTRACT Mathematical models are used to describe the relationship between two or more variables or features over the target population. Statistically, Simple Linear regression model has been extensively applied and the properties of their estimators are well known. However, this kind of model is not correctly applied in most cases, such as a longitudinal setting. Linear mixed models (LMMs) are useful when the measurements have been done over a specific interval of time. One of the most important assumptions, on both models, has been established as that the model holds for the whole data. In latter case, we could find one or several points which the function changes into. This proposal allows us to estimate the points where the model changes by minimizing a specific risk function or a loss function associated with the fitted model.