CMStatistics 2021: Start Registration
View Submission - CMStatistics
B1331
Title: Assessing timing of microrandomized trials using intensive longitudinal data and differential time-varying effect models Authors:  Nicholas Jacobson - Dartmouth College (United States) [presenting]
Kejin Wu - University of California San Diego (United States)
Abstract: In the behavioral sciences, methods for delivering interventions within the context of daily life are developing rapidly, fueled by the development of microrandomized controlled trials and experience sampling. Although intensive longitudinal data are often collected to evaluate the immediate effects of these interventions, the timing of these interventions on behavior has been given limited attention. Given this, the field could benefit from a tool to detect and estimate the time in which interventions have an impact on their respective outcomes. Nevertheless, existing tools have difficulty in estimating the timing of interventions. Consequently, we propose an extension of the Differential Time-Varying Effect Model (DTVEM) which attempts to detect the timing of interventions on outcomes by trying to detect the lag intervals between exogenous variables (i.e. intervention delivery) and outcomes. We extend the DTVEM by pairing generalized additive mixed models with linear mixed models to identify optimal time lags and intervention effects. By intensive simulations based on, the efficiency of the DTVEM with additional stage is tested, and the results showed promising power and point estimates, and low type I error. Consequently, the extended DTVEM allows researchers to perform power analyses regarding the timing of intervention effects and detect the timing of intervention effects using intensive longitudinal data and microrandomized controlled trials.