Category: Addictive Behaviors

Symposium

Exploratory Structural Equation Modeling to Improve External Validity: An Application With Substance Use Data

Friday, November 17
10:15 AM - 11:45 AM
Location: Aqua Salon E & F, Level 3, Aqua Level

Keywords: Addictive Behaviors | Psychometrics | Research Methods
Presentation Type: Symposium

At times, researchers are faced with novel research questions that require rigorous exploratory methods that allow for comparison of a range of related models examining the relations among constructs. Exploratory Structural Equation Modeling (ESEM) combines features of exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) by estimating measurement models while simultaneously estimating predictive, regression elements of the structural model. CFA requires strong theory to hypothesize the underlying factor structure, which, in practice, may not be available. Scales may not be well established, or may contain items that load onto more than one factor. These cross-loadings may be indicated either by theory or operationalization. However, CFA typically fixes cross-loadings of observed variables to zero, leading to specification of a more parsimonious measurement model than is present in the data or in the population. This fixing of cross-loadings to zero often hurts model fit, leading to the need for model modification to improve model fit, which could distort factors by over-estimating correlations between factors. ESEM allows for cross-loadings and rotates factors to improve estimation of measurement models while subjecting results to SEM model fit estimation and allows for structural predictive paths with additional constructs. A demonstration of an ESEM will be presented using data collected to differentiate sensation seeking, impulsivity, and emotion regulation as distinct constructs predicting engagement in substance use and other health risk behaviors. The demonstration will show how employing ESEM significantly improves model fit when compared to the traditional CFA approach without the need for ex post facto model modification to achieve acceptable model fit. ESEM has implications for clinical practice. By not imposing strict constraints on models our results will more readily map on to clinical experience and be more amenable to inclusion in treatment efforts.

Bradley T. Conner

Associate Professor
Colorado State University

Presentation(s):

Send Email for Bradley Conner


Assets

Exploratory Structural Equation Modeling to Improve External Validity: An Application With Substance Use Data



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Exploratory Structural Equation Modeling to Improve External Validity: An Application With Substance Use Data