Doe it matter how to fix factor means? PreviousNext
Mplus Discussion > Categorical Data Modeling >
 ffan posted on Tuesday, December 16, 2014 - 7:05 am
Hi Dr. Muthen,

I am testing measurement invariance with binary items. I followed the steps as recommended in the Mplus user guide.

When I test partial invariance, should I fix factor means at zero in one group and free in the others, or should I fix factor means in both groups at zero? For the two analyses I have done so far, I found consistent results on the modification indices for one analysis, but quite different results on the modification indices regarding which items to free for partial invariance. So it seems to me that it does matter how we fix factor means. But I do not know why? Can you explain it a little bit how Mplus works in terms of estimating factor means and why do factor means affect the modification indices provided?

I am only interested in testing configural invariance and scalar invariance, so I thought whether fixing or freeing factor means does not really matter because I am not interested in testing invariance in factor means. Same with residual variance (THETA parameterization is used in my analysis), so I fixed them all at 1 across all models I fitted in one analysis, or should I just fix the residual variance for the item being test at 1?

Thanks, Angie
 Bengt O. Muthen posted on Tuesday, December 16, 2014 - 6:36 pm
Scalar invariance involves the factor means - they are estimated in all but one of the groups. Partial scalar invariance handles the factor means the same way.
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