Combing Declarations for Count and Ce...
Message/Author
 Andre Maharaj posted on Tuesday, November 22, 2016 - 6:14 pm
Hello everyone,
I hope I am posting in the right place.

Simple question, possibly:

Can I combine the Variable declaration statements for COUNT and CENSORED?

I have a dependent variable that is of type Count, I would like to specify it as (nb), and I would like to indicate that it is censored above and below.

When I make both statements, the Output is only:

Observed dependent variables

Count
SNDSCOR3

If I remove the count declaration, I get this:

Observed dependent variables

Censored
SNDSCOR3

Here are the commands I have for the VARIABLE section:

CENSORED = sndsCor3(a) sndsCor3(b);
COUNT = sndsCor3 (nb);

Any help is appreciate!

Thanks,
Andre.
 Linda K. Muthen posted on Wednesday, November 23, 2016 - 6:14 am
You cannot combine these options.
 Andre Maharaj posted on Wednesday, November 23, 2016 - 8:35 am
Ah ok then, thanks for the response.
 Bengt O. Muthen posted on Wednesday, November 23, 2016 - 5:53 pm
Perhaps a mixture can fit this.
 Andre Maharaj posted on Wednesday, November 23, 2016 - 6:00 pm
Thanks for the suggestion!

I will try it.

I have another question related to the analysis; Instead of creating a new thread, perhaps you might suggest something here?

I am using a complex twolevel model (stratification with 8 strata and a cluster variable).

I have two count dependent measures.
I would like to combine them into a latent variable and measure the impact of a level 2 independent on the latent variable (ORF) and then reuse the same ORF variable as the outcome for the within level.

However, I get the error: "A latent variable declared on the within level cannot be used on the between level".

Any further thoughts on how I might accomplish this?

(Am I wrong in thinking that just because an observed dependent can be used on both, that a latent dependent could be used as well? If so, what am I missing?)

Thanks again!
 Bengt O. Muthen posted on Friday, November 25, 2016 - 11:09 am
You need to have an "f BY..." statement also on Between. In this case, the factor indicators are random intercepts for the count variables. You may or may not want to add equality of loadings across levels.
 Andre Maharaj posted on Friday, November 25, 2016 - 12:02 pm
Thanks for the response again.

I did add the "by" statement for both, but I used the same variable name.

I'm guessing that was the problem and I'll use another name.

However, I'm not sure I understand that last part.

Could you possibly provide a little more of an explanation? I would be really greatful.

Thanks!
 Bengt O. Muthen posted on Friday, November 25, 2016 - 3:24 pm
Yes, use different names like fw and fb.

If you have equality of loadings across levels you can view the between factor as a random intercept of the factor:

y_ij = lambda*fw_ij + lambda*fb_j + errors_ij

is the same as the multilevel expression

y_ij = lambda*f_ij + errors_ij
f_ij = fb_j + fw_ij

where fb_j is the random intercept.
 Bengt O. Muthen posted on Saturday, November 26, 2016 - 12:24 pm
Yes, use different names like fw and fb.

If you have equality of loadings across levels you can view the between factor as a random intercept of the factor:

y_ij = lambda*fw_ij + lambda*fb_j + errors_ij

is the same as the multilevel expression

y_ij = lambda*f_ij + errors_ij
f_ij = fb_j + fw_ij

where fb_j is the random intercept.