B1683
Title: A linear mixed model approach to build a calibration function from change points
Authors: Ehidy Karime Garcia Cruz - National University of Colombia (Colombia) [presenting]
Juan Carlos Correa Morales - National University of Colombia (Colombia)
Juan Carlos Salazar Uribe - National University of Colombia (Colombia)
Abstract: Linear Mixed Models have been widely studied for important authors. However, the specific approach to estimate change points subject-specific has not been worked so specifically. We present an alternative methodology to build a calibration function from change points estimated using LMMs for modeling the fixed effects and predict the random effects for each subject in a longitudinal study. The change points were estimated using evolutionary algorithms. The calibration (reverse regression) function was built under a parametric approach and it is useful to predict a change point over the time. This predicted $\hat t$ will allow to minimize an specific loss function, usually associated with the storage expenses for an specific product. The methodology will be illustrated using a real data set about dried cypress wood slats.
Journal