I'm examining a questionnaire believed to have two-factors (7 items on each variable) from previous Exploratory Factor Analysis. Looking at it using Confirmatory Factor Analysis (CFA), the fit is rather poor (RMSEA around .10; CFI around .90). So now I'm going to examine the modification indices, but I wonder whether I should:

  1. Examine the modification indices in the two-factor model, OR
  2. Examine the CFA models of each of the two variables independently first and their modification indices (and potentially remove/adjust items to fit the two independent CFA models first); and then when the two variables have a good fit examine the two-factor solution and potentially remove/adjust items.

Lastly, does anyone have any good reference for any of these approaches.

Any help is much appreciated, Gorp.


1 Answer 1


I think it is best to go with alternative 2. as this examines the unidimensionality of each scale. See Ziegler and Hagemann (2015; Testing the Unidimensionality of Items: Pitfalls and Loopholes; European Journal of Psychological Assessment) stating:

"start... with testing the measurement models for each new scale score..., followed by testing structural models including measurement models for overlapping traits..., should provide much better evidence with regard to unidimensionality."


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.