Invited Symposium
The State-of-the-Science on Factorial Invariance
Sunday, May 24, 2009,
1:00 PM - 2:20 PM
Yerba Buena 3 - 4
Chair:
Todd D. Little
University of Kansas |
Factorial invariance involves testing the assumption that measures of constructs have the same meaning in two or more groups or two or more measurement occasions. Tests of construct differences can only be examined after the comparable meaning of the constructs is established. We present state-of-the-science thinking on this essential concept.
Factorial Invariance: What It Is and Why It Is So Damn Important
Todd D. Little
University of Kansas
I will provide a conceptual overview of what factorial invariance is and why we need to test for it. I'll draw parallels to Differential Item Functioning and other forms of testing for whether our measures are comparable across one or more groups or two or more time points.
Prediction Using Measures That Violate Factorial Invariance: What Happens?
Roger Millsap
Arizona State University
For measures that fit a common factor model in multiple groups, the question of invariance
in the factor structure across groups (i.e., factorial invariance) can be explored empirically.
When violations of invariance are found, and we want to use those measures in prediction,
what are the consequences? This question is particularly important in the use of measures for
selection. The question is addressed here for linear prediction.
Time For Change?: Redefining and Reconsidering the Role of Factorial Invariance in Psychological Inquiry
Nilam Ram
Pennsylvania State University
Dominant methods for assessing factorial invariance invoke a set of beliefs and assumptions about the nature of data that seems somewhat at odds with the basic tenants of psychological inquiry. Although psychological science focuses on studying the behavior of individuals, the methodological procedures used often ignore or even suppress idiosyncratic aspects of behavior that may interfere with the development of general laws. Therefore, we suggest that the definitions of invariance be reconsidered and that new conceptions and approaches be developed. Using simulations and empirical data we present and illustrate the potential utility of conceptual and methodological frameworks that explicitly allow for an integration of individual behavioral realities and scientific need for universal laws. We underscore the need for precision in both the theoretical conceptions and methodological invocations of when and how individual and invariant aspects of behavior can and may be observed.
Co-Author:
John R. Nesselroade, University of Virginia
Keith F. Widaman (Discussant)
University of California, Davis
