Quantitative Methods Forum

When:
January 23, 2012 @ 1:00 PM – 2:00 PM
2012-01-23T13:00:00-05:00
2012-01-23T14:00:00-05:00
Where:
Norm Endler Seminar Room (BSB 164)
Cost:
Free

Speaker: Dr. Dave Flora & Phil Chalmers, York University
                 Department of Psychology

Title: Strange Signs in Confirmatory Factor Analysis: Illustration Using ADHD Data.
 
Abstract: We (e.g., Toplak, Pitch, Flora, et al., 2009) have consistently found that bifactor models explain the covariation among the 18 symptoms of Attention-Deficit/Hyperactivity Disorder as well as or better than several competing factor models. That is, we obtain this result across a variety of samples (children, adolescents, undergraduates, adults) and a variety of measures (semi-structured clinical interview vs. rating scale, self-report, parent-report, teacher-report). Yet, the fitted models usually have a curious feature: Even though all of the observed correlations are moderately to strongly positive, a number of specific-factor loadings have negative signs. Although this feature poses a challenge for conceptual interpretation, at least we can explain why it occurs mathematically. Moreover, the signs of the specific-factor loadings may flip depending on how model identification constraints are imposed. We demonstrate these results using the bifactor model, but they could also easily occur with other kinds of confirmatory factor analysis models, such as multitrait-multimethod models.

Presentation Slides