Cluster or multi-level CFA? PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Melanie posted on Tuesday, April 02, 2013 - 11:09 am
I want to examine the factor structure of a questionnaire amongst a group of raters. The raters are clustered into groups based on who they have rated. I’m not interested in the ratees' scores at all at this point. I just want to know if the questionnaire factors well across the raters, and reviewers have told me I need to take into account the fact that they are grouped based on who they have rated.

So I’m confused as to whether I’m ok just using the cluster and type=complex or whether I need to use multi-level modelling and specify within and between models (which presumably will be the same as I’m not looking at/entering the ratee data at this point).
 Linda K. Muthen posted on Tuesday, April 02, 2013 - 12:23 pm
How many raters do you have?
 Melanie posted on Wednesday, April 03, 2013 - 1:56 am
I have 351 raters, and the number of clusters is 110 (so I'm guessing that means that there are 110 ratees with ratings), and the average cluster size is 3.19.
 Linda K. Muthen posted on Wednesday, April 03, 2013 - 2:17 pm
Do you have 351 ratees and 110 raters. If not, what is cluster? Aren't ratees nested in raters such that you would have more ratees that raters?
 Melanie posted on Wednesday, April 03, 2013 - 2:34 pm
I have 110 people (ratees) who have been rated by a total of 351 raters.

I am clustering on the the person they rated (so Rater 1, 2 and 3 might rate ratee A. Rater 4, 5, 6 and 7 might rate ratee B and so on - the raters never rate more than one person)

So the average cluster size (no. of raters who rate one person) is 3.19

and I'm just looking to see if the questionnaire factors well across the rater's responses, taking into account that several raters may be rating the same person on that questionnaire. I don't want to look at ratee responses on that questionnaire.

Its essentially a 360 degree feedback process and I just want to look at whether the 360 version of the questionnaire factors well. I'm not interested in the self-rated version of the questionnaire.

Sorry if this is confusing!
 Linda K. Muthen posted on Wednesday, April 03, 2013 - 4:46 pm
I would use TYPE=TWOLEVEL and perhaps TWOLEVEL EFA to get an idea of the factor structure on the two levels with CLUSTER=ratee.
 Melanie posted on Thursday, April 04, 2013 - 1:57 am
Thank you, I'll give it a whirl!
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: