Modeling the psychometric properties of information manipulation ratings

Abstract

This paper provides a reanalysis of two large data sets (Jacobs, Dawson & Brashers, 1996a; McCornack, Levine, Solowczuk, Torres, & Campbell, 1992) involved in a recent dispute over the validity of Information Manipulation Theory (IMT) (McCornack, 1992; McCornack et al., 1992). McCornack, Levine, Morrison and Lapinski's (1996) results were consistent with a second‐order factor model predicted by IMT while Jacobs, Brashers and Dawson (1996b) found that the same data were first order multidimensional with quality and honesty ratings loading on the same factor. Because the results of exploratory factor algorithms, especially when coupled with varimax rotations, are suspect when second‐order unidimensionality present, the Jacob et al. (1996b) results are dubious, and published errors may exist. The reanalyses reported here demonstrate that the second order model predicted by IMT provides an excellent fit to the data, while the first order model advanced by Jacobs et al. does not. The curren t results provide further evidence that the use of exploratory factor analysis (EFA) with varimax rotations produce erroneous results in the context of second‐order unidimensionality.