Classification of Longitudinal Data Using P-Splines and Correlated Mixed-Effects Models Applied to Predict Pregnancy Outcomes

• Activity Number: 542
• Type: Contributed
• Date/Time: Wednesday, August 3, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Cristian Meza* and Ana Arribas-Gil and Rolando De la Cruz and Claudio Fuentes
• Companies: Universidad de Valparaiso and Universidad Carlos III de Madrid and P. Universidad Católica de Valparaíso and Oregon State University
• Keywords: Semiparametric ; P-Splines ; Mixed effects models ; Correlated errors

Abstract

We propose a semiparametric mixed-effects model based on Splines method to the classification problem for longitudinal data. The motivation of this work is the study of a real dataset of hormone levels measured at the early stages of pregnancy that can be used to predict normal versus abnormal pregnancy outcomes. The proposed model is a particular case of the semiparametric nonlinear mixed-effects class of models (SNMM) in which finite dimensional (fixed effects and variance components) and infinite dimensional (an unknown function) parameters have to be estimated. The unknown function is estimated in a nonparametric fashion using a Penalized Splines procedure. We compare the advantage on the use of random effects versus the modeling of the correlation in the errors. For this specific dataset, this method outperforms previous results using both parametric and semiparametric mixed effects models. Moreover, the obtained results with this method support the idea that the modeling of the serial correlation on the error terms may compensate the lack of random effects.