Exploring Alternative Factor Structures for the Colorado Learning Attitudes about Science Survey

The Colorado Learning Attitudes about Science Survey (CLASS) probes respondents' attitudes on learning physics. The CLASS consists of 42 total items (36 total which are scored), organized into eight categories. These eight categories contain 26 items total, with several items appearing in more than one category. Douglas et al. (2014) conducted a factor analysis study on the CLASS, and proposed a simplified factor structure for the instrument, consisting of three factors and 15 items, with no items loading on multiple factors. The current study examines the Douglas results on a different dataset. This dataset comes from a large eastern US land-grant institution and focuses on post-test CLASS responses from the introductory calculus-based mechanics course. Initial confirmatory factor analysis (CFA) results indicate that neither the Douglas factor structure nor the originally published categories fit the data for this study. This study then looks at applying the exploratory factor analysis (EFA) method employed by Douglas to this new dataset to analyze differences between the two studies. Differences were identified in initial EFA results; the manner in which the data was split also impacted the results. Results also varied depending on the cutoff values used in the EFA process. This presentation will discuss the process used, variations found between results, and the application of CFA to the EFA results. Finally, the identification of subscales and unique items will be highlighted.