Conferences

In this talk we will compare a hierarchical continuous-time autoregressive model with more traditional modeling approaches such as the linear mixed model (LMM) framework for analyzing intensive longitudinal data. Although both frameworks have been applied for such designs and they both model individual differences, their relation is unclear and has not been thoroughly investigated yet. Also, while the estimation in the LMM framework is quite straightforward and implemented in many general purpose and specific software packages, the estimation of continuous-time autoregressive models can prove relatively cumbersome. We will focus on one particular continuous-time autoregressive model, namely the hierarchical Ornstein-Uhlenbeck (OU) process with measurement error. Our aim is to introduce this dynamical model by interpreting its parameters as related to the mixed models framework. First we will show that under certain conditions, the OU model is equivalent to a linear mixed model. Next, we will demonstrate under which conditions the equivalence relation breaks down. We will conclude that the OU model based dynamical approach is especially relevant when we can assume several sources of inter-individual differences.