E/CFA with indicators having strong c... PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
 Lois Downey posted on Friday, March 20, 2015 - 1:41 pm
I want to do an E/CFA on a large group of indicators that are scored 0-10 and have strong ceiling effects. Because modification indices are needed for E/CFA, I can't define the indicators as censored from above and base the analysis on tobit regression. The only two options I can think of are to (1) ignore the ceiling effect and define the indicators as continuous; or (2) reduce the number of response categories to 10 or fewer (perhaps collapsing responses 0 and 1 into a single category) and define the indicators as ordered categorical. Which, if either, of these two options would you recommend? Are there other options I have overlooked that would be preferable?

 Bengt O. Muthen posted on Friday, March 20, 2015 - 4:07 pm
You get modindices with censored variables using the WLSMV estimator. You can also reduce the number of categories, collapsing some, and treat the variables as ordered categorical.

A more advanced approach if you really care about the ceiling effect part of the variable is 2-part FA.
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