Strange output for residual variances
Message/Author
 Richard E. Zinbarg posted on Tuesday, September 05, 2006 - 1:42 pm
Hi Linda and/or Bengt,
A student and I are running an analysis in which we have 2 factors (neuroticism and depression) each indicated by multiple indicators with 1 factor regressed on the other one (depression being regressed on neuroticism). Mostly everything looks good in the model output except for the residual variances for two of our indicators (GENDEP and ANDEP) which have asterisks in the 2nd to last column (std). The unstrandardized residual variances look reasonable and the fully standardized (stdyx) resicual variances look reasonable. Do the asterisks indicate a problem with the model or can we safely ignore them? I am pasting our strange residual variance output below in case that will be helpful.
Thanks!
Residual Variances
GENDEP 1568.375 161.519 9.710 ******* 0.219
ANDEP 2086.853 139.738 14.934 ******* 0.490
D 1.000 0.000 0.000 0.255 0.255
 Linda K. Muthen posted on Tuesday, September 05, 2006 - 3:04 pm
The asterisks mean that the value could not be fit in the space allocated. It looks like your variables have large variances. I suggest rescaling. We recommend keeping variances between 1 and 10.
 Richard E. Zinbarg posted on Tuesday, September 05, 2006 - 7:04 pm
many thanks for the quick reply!
 Chun-Yi Lee posted on Tuesday, April 24, 2012 - 11:47 pm
Dear Dr. Muthen,

I have tried conducting path analysis using Mplus.
However, it is strange that "Est./S.E." for the residual variances are all the same.
The output is as follows.
--------------------------------------------------------------
Residual Variances Estimates S.E. Est./S.E.
x1 9.298 0.299 31.081
x2 13.629 0.439 31.081
x3 2.699 0.087 31.081
x4 0.057 0.002 31.081
y1 24.355 0.784 31.081
y2 11.421 0.367 31.081
---------------------------------------------------------------

I would like to ask some questions.
Q1: Is there any upper bound for "Est./S.E."?
Yet, it is still strange since 13.629 / 0.439 = 31.046 < 31.081?
Q2: I was wondering why "Est./S.E." is not equal to Estimates divided by S.E.?

Thank you so much for your time.
 Linda K. Muthen posted on Wednesday, April 25, 2012 - 10:55 am
Please send the output and your license number to support@statmodel.com.