Message/Author |
|
Arpana posted on Thursday, August 10, 2006 - 7:55 am
|
|
|
Dear Dr. Muthen, I am trying to fit a FMA with 2 latent classes (where one is a zero class) and each class has 2 factors: MODEL: %overall% drug1 by sed* cann@1 coke tranq opit amphe hall inhal@0 ; drug2 by sed@0 cann@1 coke tranq opit amphe hall inhal; drug1 with drug2 ;drug1*1; drug2*1;[drug1@0] ;[drug2@0] ; %c#1% [cann$1-inhal$1@15] ; drug1 by cann-inhal@0 ; drug2 by cann-inhal@0 ; drug1@0 ; drug2@0 ; drug1 with drug2@0 ; %c#2% [cann$1-inhal$1] ; drug1 by sed cann@1 coke tranq opit amphe hall inhal@0 ; drug2 by sed@0 cann@1 coke tranq opit amphe hall inhal ; drug1 with drug2 ;drug1 ; drug2 ; I get several error messages related to model identification. Is there a parameter/s I should constrain to identify this model? |
|
|
I can't see a problem off hand, so please send your input, output, data and license number to support@statmodel.com. Be sure to include Tech1 in the output. |
|
Arpana posted on Monday, August 14, 2006 - 6:11 am
|
|
|
Dear Dr. Muthen, I fixed the factor variances in c#2 to 1.0 and that seemed to do it. Thanks! |
|
|
But you already have one loading fixed at 1 for each factor, which identifies the factor variances. Seems like there is something else going on. You have a regular factor model in class 2 with factor means zero so that should be fine. Look at Tech1 to see that you get the parameters freed that you expect. |
|
Arpana posted on Thursday, August 24, 2006 - 6:37 am
|
|
|
Dear Dr. Muthen, I have emailed the code and data to the above email address and look forward to your suggestions. |
|
Back to top |