Estimation Methods PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
 Abdullah Ali posted on Tuesday, November 27, 2018 - 10:09 pm
Dear Dr Muthen,

I would like to run a CFA and SEM with ordinal variables measured on a 1-7 Likert scale and a categorical variable (with 5 categories).

Mardia’s test indicates that the data are not multivariate normal. My questions concern estimation methods and approaches to modelling.

1. Do I run a CFA with both ordinal and categorical variables (using the WLSMV estimator) or

2. Can I a run a CFA with just the ordinal variables (the categorical variable is excluded) using the MLR estimator and then with the structural model, adding back in the categorical variable, run a full model (with both ordinal and categorical variables) using the WLSMV estimator.

Many thanks,

 Bengt O. Muthen posted on Wednesday, November 28, 2018 - 4:58 pm
When you say "categorical" as opposed to ordinal, perhaps you mean nominal. If so, you can't use WLSMV but need to use ML.
 Abdullah Ali posted on Thursday, November 29, 2018 - 9:21 pm
Thank you for your reply!

The categories are for academic grades as follows:

1 = Fail

2 = Pass

3 = Credit

4 = Distinction

5 = High Distinction

(higher numbers align with higher grades)

I am just not sure if these should be treated as categorical or ordinal or something else.

Best Regards,
 Bengt O. Muthen posted on Friday, November 30, 2018 - 4:41 pm
You can treat them as ordinal, that is, put it on the Categorical list.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message