2

I am learning currently CFA after completing EFA using Andy Fields book and Using Multi-variate statistics by Barbara Tabachnick and Linda Fidell. In EFA the assumptions of the test are mentioned as

  • Adequate Sample Size
  • Normality of Data
  • Linearity
  • Outliers
  • Multi-collinearity

Because CFA is basically the same thing with constraints (setting loadings to zero based on pre-defined theory), do the same assumptions need to be tested first?

Thank you for your time

John Smith
  • 215
  • 1
  • 4
  • 19
  • 3
    Excuse for a moment, your list of assumptions of EFA "test" (sic - ??) is partly confusing. What is about "multi-collinearity" - should it be or not be? Also, Normality of data is not always needed. Or are you speaking specifically of testing statistical hypotheses (which are indeed the core in CFA and sometimes done in EFA)? EFA assumptions I've listed [here](http://stats.stackexchange.com/a/198684/3277). – ttnphns Jan 15 '17 at 13:18
  • 1
    Generally speaking exploratory factor analysis is intended to be descriptive and help lead you to formulating statistical tests. In confirmatory factor analysis you formally apply those statistical tests. – Michael R. Chernick Jan 15 '17 at 15:32
  • Hi @ttnphns, apologies should have been more clear. For EFA from my understanding the data should have some multi-collinearity because i guess thats whats being tested in the end. In relation to your link. If you have a data set with many constructs and you are implementing CFA based on a theme so taking a subset of constructs and running CFA. Is it good practice to run through the assumptions again for these subset of answers or will testing the assumptions once on the overall dataset suffice? – John Smith Jan 15 '17 at 15:35
  • Logically, assumptions pertain only to data (items) which participate in this or that CFA. – ttnphns Jan 15 '17 at 16:04

0 Answers0