Design, measurement, and analysis considerations and evaluations in intensive longitudinal method

Abstract

This dissertation examined the accuracy of a few selected statistical approaches in evaluating invariance in measurement and mediation with the presence of planned-missing data in the context of intensive longitudinal method (ILM). The planned-missing data design was implemented as a three-form design where a portion of measurement scale items were selectively removed for individuals at each measurement occasion with the purpose of reducing participation burden and fatigue stemming from the burst of measurements in ILM ranging from 2 to 12 measurement occasions per day. Three simulation studies were conducted with the aim of providing insights and recommendations to applied researchers in the design, measurement, and analysis of intensive longitudinal data. Study 1 compared two methods for testing intensive longitudinal measurement invariance in their performance in detecting invariant and non-invariant measurement parameters. Study 2 and Study 3 evaluated the performance of the dynamic structural equation model (DSEM) framework in estimating the time-invariant and time-varying effects of longitudinal mediation models. Sample sizes (N), length of measurement occasions (T), percentage of planned-missing data (PMD), and effect sizes were manipulated in the simulation studies. The dissertation concluded with recommendations for applied researchers. Limitation and areas for future research were also discussed.