Investigating the Nature of Method Factors Through Multiple Informants: Evidence for a Specific Factor?

Abstract

When a self-report instrument includes a balanced number of positively and negatively worded items, factor analysts often use method factors to aid model fitting. The nature of these factors, often referred to as acquiescence, is still debated. Relying upon previous results (Alessandri et al., 2010; DiStefano & Motl, 2006, 2008; Rauch, Schweizer, & Moosbrugger, 2007), we submit that the so-called method factors, instead, represent substantive specific factors. This study investigates the convergence of method effects across different observers. The revised Life Orientation Test (Scheier, Carver, & Bridges, 1994) was administered to a sample of 372 adults (57% females), with 372 acquaintances serving as informants. Results showed that a specific factor was detectable both with self- and other-ratings. A significant correlation across informants provided evidence for the convergence of this specific factor. Construct validity was examined by locating this specific factor within a nomological net of personality variables. Theoretical and practical implications of the findings are discussed.

Notes

¹Goodness-of-fit was evaluated using the chi-square, the root mean square error of approximation (RMSEA) with a 90% confidence interval, the Tucker and Lewis fit index (TLI), and the Comparative Fit Index (CFI). Multiple indices were selected as they provide different information for evaluating model fit. Hu and Bentler (1998) suggested the following cutoff values for the preceding indexes: RMSEA ≤ .06, TLI ≥ .95, and CFI ≥ .95.

Note. ^a Correlation between factors was −.53. CFI = Comparative Fit Index; TLI = Tucker and Lewis fit index; RMSEA = root mean square error of approximation.

Note. ^a Correlation between factors was −.56. CFI = Comparative Fit Index; TLI = Tucker and Lewis fit index; RMSEA = root mean square error of approximation.

²The correlation matrix is available from the corresponding author.

³Preliminary analyses were performed using the Correlated-Traits-Correlated Method (CTCM) approach. A necessary condition for the typical CTCM model to be identified is the presence of at least three traits and three methods (Marsh, 1993). With only two methods, the measurement model of the current study is not identified and is therefore not estimable. In order to achieve identification, we imposed additional restrictions on the loadings of the method factors. However, serious difficulties were encountered during estimation, which resulted in a number of improper solutions. Given these results, the Correlated Uniqueness model was preferred to the CTCM model.

*p < .05.

**p < .01.

⁴In this section, we not discuss further results related to the optimism substantive factor because they are not of interest with respect to the aims of this article.