Question regarding example penn-app6
Message/Author
 Anonymous posted on Friday, May 07, 2004 - 11:21 am
In this example, you have the following syntax:

f1a BY AUDY1@1 AUDY2@1 AUDY3@1 AUDY4@1 AUDY7@1;
f2a BY AUDY1@0 AUDY2@1 AUDY3@2 AUDY4@3 AUDY7@6;

f1t BY tdy1@1 tdy2@1 tdy3@1 tdy4@1 tdy7@1;
f2t BY tdy1@0 tdy2@1 tdy3@2 tdy4@3 tdy7@6;

[audy1\$1-tdy7\$1@0];

Can you help me to understand what the last line is doing? And, alternatively, could you have estimated the threshold but held them equal for all items...e.g. [audy1\$1-tdy7\$1*0] (1);

I seem to recall this being done in other binary growth examples. Thank you!
 Anonymous posted on Friday, May 07, 2004 - 11:32 am
One other question to follow up, when one refers to the growth factors (f1a, f2a, etc.), should they be considered in the metric of log odds. For example, if you had a class that had no probability of AUDY at Y1, would you contstain f1a to be -15 (the threshold limit)?
 Linda K. Muthen posted on Friday, May 07, 2004 - 12:08 pm
For binary outcomes, I would specify the model as follows:

f1a BY AUDY1@1 AUDY2@1 AUDY3@1 AUDY4@1 AUDY7@1;
f2a BY AUDY1@0 AUDY2@1 AUDY3@2 AUDY4@3 AUDY7@6;

f1t BY tdy1@1 tdy2@1 tdy3@1 tdy4@1 tdy7@1;
f2t BY tdy1@0 tdy2@1 tdy3@2 tdy4@3 tdy7@6;

[audy1\$1-tdy7\$1] (1);

[f1a@0 f1t@0 f2a f2t];

{audy1@1 uudy2-sudy7 tdy1@1 tdy2-tdy7};

for weighted least squares using the Delta parameterization.

The growth factors are continuous variables.
 Anonymous posted on Saturday, May 08, 2004 - 12:32 am
Thank you, I overlooked the fact that the penn example had continuous indicators. I applied the syntax that you suggested and I get an error message telling me that syntax in the curly brackets (refering to the scale factors) is ignored. I can get the model to run if I do not run a mixture model (that is, if I specify just a binary growth model). I'm using version 2.
 Linda K. Muthen posted on Saturday, May 08, 2004 - 6:13 am
Why don't you send your output where you get the error message to support@statmodel.com. I need to see your entire input to answer your question.
 Linda K. Muthen posted on Saturday, May 08, 2004 - 9:53 am
Thanks for sending the output. In mixture modeling, you don't need the scale factors.