Message/Author |
|
Daniel posted on Thursday, February 20, 2003 - 7:49 am
|
|
|
What is the problem that StdYX exceeds one? Does that matter? If StdYX (or Std) is greater than 1, e.g.,1.020, the residual variance will be -0.040. And R-Square is undefined. Following that there is 0.10400E+01. Should I use 0.10400E+01. as R-Square? |
|
bmuthen posted on Thursday, February 20, 2003 - 8:42 am
|
|
|
If this is a 1-factor model you have an inadmissible solution (since variances should be positive) and you would want to modify your model. |
|
daniel posted on Thursday, February 20, 2003 - 11:08 am
|
|
|
This is five-factor model |
|
bmuthen posted on Thursday, February 20, 2003 - 11:12 am
|
|
|
This may be a sign that the factor covariance matrix is not positive definite as it should be. For example, check for correlations greater than 1. However, for a fuller treatment of the topic of standardized coefficients greater than 1, see Joreskog's writing at www.ssicentral.com/lisrel/column2.htm |
|
sivani sah posted on Monday, April 03, 2006 - 6:14 am
|
|
|
Thank you.I am sorry, this site is not working. |
|
|
Try www.ssicentral.com and go to Karl's corner. Perhaps they have changed the link. |
|
Tim Cupery posted on Monday, September 10, 2012 - 11:55 am
|
|
|
here is the current link to the Jöreskog paper, in case anyone happens across this thread: http://www.ssicentral.com/lisrel/techdocs/HowLargeCanaStandardizedCoefficientbe.pdf |
|
Back to top |