Number of Subjects and Time Points Needed for Multilevel Time Series Analysis: A Monte Carlo Study of DSEM in Mplus Version 8.
Authors: Mårten Schulzberg (presenter), Bengt Muthén
Dynamic Structural Equation Modeling (DSEM) provides new methods for analyzing intensive longitudinal data such as that obtained with ecological momentary assessments, experience sampling methods and ambulatory assessments. DSEM uses two-level modeling with time on level 1 and individuals on level 2. It models intra-individual changes over time and allows the parameters of these processes to vary across individuals using random effects. This is made possible using Bayesian estimation. As an extension of conventional multilevel modeling, DSEM allows random effects to not only be dependent variables regressed on level 2 covariates but also allows them to be predictors of various outcomes. There are three key random effects of interest in psychological research of longitudinal data, random mean, random autocorrelation and random residual variance. A series of increasingly more complex models of this kind are
used in a Monte Carlo study that varies several factors: Number of subjects, number of time points, variance explained in dependent variables, intra-class correlation and model misspecication. Preliminary results indicate that it is more diffcult to obtain good estimates for models using the random effects as predictors. Practical considerations of Monte Carlo simulations using Bayes are