Cluster on FactorScores v.s Individua... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Michael G posted on Friday, May 18, 2012 - 3:17 am
Is there any statistical issue if we use factor scores rather than individual variables in cluster analysis? Your input would be greatly appreciated
 Linda K. Muthen posted on Friday, May 18, 2012 - 12:06 pm
Using a factor with individual variables as indicators versus using factor scores is not the same thing unless factor score determinacy is one. This is seldom if ever the case.
 Michael G posted on Thursday, May 31, 2012 - 4:15 am
Thanks Linda for your insightful input. I haven't thought before about determinacy, which is interesting!

I do have continuous high dimensional variables which they are highly correlated. I want to do dimension reduction using PCA or factor analysis. Then using either the highest loading factors variables or the factor score to construct clusters. I tried about four approaches (using all the variables; highest loading variables; principal component scores, and factor scores) but the clusters I found using the factor scores makes more sense in regarding to my research purpose. Would you please kindly suggest me if there is any other plausible approaches.
Thanks Michael
 Linda K. Muthen posted on Thursday, May 31, 2012 - 11:40 am
The only other thought is a multiple indicator growth model.
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: