I'm trying to perform a CFA on data where all of the observable variables are categorical. I believe I have entered the data correctly - the summary of the proportions across categories are correct but the rest of the sample statistics are causing concern. Can you clarify what exactly the means/intercepts/thresholds are? Also the correlation matrix is completely different to the one obtained in Stata. Both appear to be calculated on the same number of observations.
The sample statistics for the default weighted least squares estimator for a CFA model are thresholds and tetrachoric or polychoric correlations. See the Topic 2 course handout and video for a discussion of categorical data analysis.
Luning Sun posted on Monday, March 17, 2014 - 10:56 am
I have a dataset with around 200 participants. The test has 10 dichotomous items loading on one factor. All of the items are extremely easy, with low frequencies of incorrect responses. When I run CFA analysis, most of the unstandardized factor loadings are not significant, however, the standardized (STDYX)factor loadings turn significant. I am very confused with this result. Can you help me understand it please? Thank you very much!