Reliability Assessment for Multivariate, Multilevel Intensive Longitudinal Data with Integrated Trends and Intraindividual Variability.

Add to Calendar	Outlook Web Yahoo Google Download .ics
Time	Wed, Oct 8, 2025 11:00 am to 12:00 pm
Location	HHD 101 Google map
Presenter(s)	Our speaker for this week is Xiaoyue Xiong, Doctoral student in the Department of Human Development and Family Studies (HDFS) at Penn State.
Description	Recent tools for evaluating reliability in dynamic structural equation models (DSEMs) offer a framework for assessing reliability in multivariate, multilevel intensive longitudinal data. In the first section, I will discuss issues and estimation results from fitting multivariate Nonlinear Growth Curve models with Vector AutoRegression (VAR) in the residuals, and random effects in both the trends and VAR components. I will address how reliability can be computed separately for the trend and VAR components, and implications of ignoring trends on reliability assessment. In the second section, I will present a research proposal for developing a tool for reliability and power analysis that can accommodate irregularly spaced, multivariate, multilevel intensive longitudinal data. The proposed simulation studies will address three key questions: which items show differential between- versus within-person reliability for optimal screening and monitoring, how reliability varies across temporal scales to inform measurement schedules, and whether reliability patterns differ across individual characteristics.

Add to Calendar

Outlook Web Yahoo Google Download .ics

Time

Wed, Oct 8, 2025 11:00 am to 12:00 pm

Location

HHD 101

Presenter(s)

Our speaker for this week is Xiaoyue Xiong, Doctoral student in the Department of Human Development and Family Studies (HDFS) at Penn State.

Description

Recent tools for evaluating reliability in dynamic structural equation models (DSEMs) offer a framework for assessing reliability in multivariate, multilevel intensive longitudinal data. In the first section, I will discuss issues and estimation results from fitting multivariate Nonlinear Growth Curve models with Vector AutoRegression (VAR) in the residuals, and random effects in both the trends and VAR components. I will address how reliability can be computed separately for the trend and VAR components, and implications of ignoring trends on reliability assessment.

In the second section, I will present a research proposal for developing a tool for reliability and power analysis that can accommodate irregularly spaced, multivariate, multilevel intensive longitudinal data. The proposed simulation studies will address three key questions: which items show differential between- versus within-person reliability for optimal screening and monitoring, how reliability varies across temporal scales to inform measurement schedules, and whether reliability patterns differ across individual characteristics.