Pairwise estimation of correlation PreviousNext
Mplus Discussion > Structural Equation Modeling >
 Daniel Thurm posted on Tuesday, February 28, 2017 - 3:31 am
Dear all,

I have 10 latent variables, each consisting of 5 indicators.
I am interested in the correlation between the latent variables. The sample size is only 200 so I conclude that I can not estimate the whole model at once. But would it be possible to estimate the correlation pairwise, which means I would always take 2 latent variables and estimate the correlation between them. And then continue to do this for all 45 pairs?

It would be glad to get some help on this issue.

Best wishes!
 Bengt O. Muthen posted on Tuesday, February 28, 2017 - 6:12 pm
I think you can analyze all 50 variables and 10 factors in one analysis. It is also a better treatment of missing data.
 Daniel Thurm posted on Wednesday, March 01, 2017 - 8:53 am
Thank you very much for your answer! It is usually recommended to have a sample size 5 times bigger than the number of parameters that are estimated. Is there an literature that says that a smaller sample size is enough?

In addition I have a another data set where, if I include all latent variables, the number of parameters that are estimated is bigger than the sample size. Would you then recommend a pairwise estimation of correlation between the latent variables or is this generally not a good way?

Thanks a lot!

Best wishes, Daniel
 Bengt O. Muthen posted on Wednesday, March 01, 2017 - 9:24 am
You are right in your sample size recommendation, but I don't think you buy anything by doing it pair-wise.
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