Message/Author 

Xu, Man posted on Thursday, September 11, 2008  4:09 pm



I used 25 multiply impuated data sets for an analysis with MI. In the oupt I got 24 computated succesffully instead of 25. What should I do? Thanks! 


If you add TECH9 to the OUTPUT command, you will get a message about the problem with the one data set. 

Xu, Man posted on Friday, September 12, 2008  7:47 am



Thanks! It's really good to know. I'll run the analysis again adding this request. BTW: is it that in a multilevel SEM, if I have a latent dependent variable, then I have to also specify a level two factor for it, if I have group level predictor? Thanks! 


Yes. 

Xu, Man posted on Sunday, September 14, 2008  10:28 am



Thank you! I have checked the output of tech9. Apparently although only 1 out of the 25 data didn't give successful computation, 5 other datasets gave warning messages like WARNING: THE RESIDUAL COVARIANCE MATRIX (THETA) IS NOT POSITIVE DEFINITE THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR AN OBSERVED VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO OBSERVED VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO OBSERVED VARIABLES. CHECK THE RESULTS SECTION FOR MORE INFORMATION. 

Xu, Man posted on Sunday, September 14, 2008  10:30 am



And the 1 failed dataset has this warning: THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO AN ILLCONDITIONED FISHER INFORMATION MATRIX. CHANGE YOUR MODEL AND/OR STARTING VALUES. THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO A NONPOSITIVE DEFINITE FISHER INFORMATION MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS 0.488D17. THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THIS IS OFTEN DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. CHANGE YOUR MODEL AND/OR STARTING VALUES. PROBLEM INVOLVING PARAMETER 37. In this case do you think I can go ahead using results from the 24 datasets only or try to run analysis for each of the data to find out what's the problem? But if I have a remedy for each data, it woulnd't be multiple imputation anymore... is it? Thanks! 


You should run all of the data sets separately that you received warnings for to see what the problem is. None of these messages should be ignored. If you have further questions on this, please send your input, data, output, and license number to support@statmodel.com. 

Back to top 