R-SQUARE
Message/Author
 MBH posted on Thursday, May 01, 2014 - 10:56 am
Dear Linda,

There is something which puzzles me regarding R-SQUARE for SEM estimated by weighted least square.

In the output under R-SQUARE, Mplus is only producing residua variance for some variables but not all. Is that normal or does this mean three is a problem with the data? For example:

R-SQUARE
Observed Variable - Estimate - Residual Variance

Var1 - 0.080
Var2 - 0.677 - 0.168
Var3 - 0.235 - 0.154
Var4 - 0.869 - 0.342
Var5 - 0.158 - 0.674
Var6 - 0.124 - 0.629
Var7 - 0.550
Var8 - 0.030
Var9 - 0.480 - 0.689

Thank you
 Linda K. Muthen posted on Thursday, May 01, 2014 - 11:49 am
The residual variances for the continuous variables are part of the regular results. Residual variances are not model parameters for categorical variables. The residual variances shown with R-square are computed after model estimation as remainders.
 MBH posted on Thursday, May 01, 2014 - 4:34 pm
Thanks Linda,

Another question, if a CFA model has some residual errors missing from the diagram, can that be problematic even though Mplus identifies the model?

Say a CFA is measured by 10 variables and the diagram only shows residual arrows for 6 and not for the remaining 4, is that something to be concerned about?

Thanks
 Linda K. Muthen posted on Thursday, May 01, 2014 - 5:24 pm
Are the indicators for which there are no residual arrows categorical?
 MBH posted on Thursday, May 01, 2014 - 5:29 pm
Hi Linda,

Some are categorical and some are continuous, which look like a ordered categorical scale
 Linda K. Muthen posted on Thursday, May 01, 2014 - 8:08 pm
I am trying multilevel analysis everything is ok but I am not getting value of R-Square. I do not know what is the problem. please guide.

analysis: type = twolevel random;
model:

%within%
il on gen,edu,corgexp,cmwork,texp,lmx,cvoice,svoice,sfocus,
jfocus,lmxxcv,lmxxsv,lmxxsf,lmxxjf;
%between%
il;

output: stdyx ;
 Bengt O. Muthen posted on Wednesday, March 04, 2015 - 3:29 pm
With Type=Random you don't get R-2. This is because Type=Random expects a random slope which makes the variance of the DV vary as a function of the IV. I don't see that your model needs Type=Random.