Two-part model with an ordinal outcome PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Mplususer  posted on Tuesday, December 12, 2006 - 10:19 pm
Dear Dr. Muthen,

Would you mind dropping me a hint about modeling two-part GMM with an ordinal outcome. My question is as follows:
**********************************************************************
Let's use Example 6.16 in Mplus user's guide

DATA: FILE = ex6.16.dat;

DATA TWOPART:
NAMES = y1-y4;
BINARY = bin1-bin4;
CONTINUOUS = cont1-cont4;
! Can it be changed to Categorical = cont1-cont4 for my ordinal outcome?

VARIABLE: NAMES = x y1-y4;
USEVARIABLES = bin1-bin4
cont1-cont4;
! Do I need to add
Categorical = cont1-cont4 here ?

CATEGORICAL = bin1-bin4;
MISSING = ALL(999);
! Can I add Classes = (2) for
mixture modeling
 Mplususer  posted on Tuesday, December 12, 2006 - 10:20 pm
ANALYSIS: TYPE = MISSING;
! Can I specify Type = Mixture Missing; here?
ESTIMATOR = MLR;
MODEL: ! Can I add %Overall% and %c#1% statements for my two-part growth mixture model
iu su | bin1@0 bin2@1 bin3@2 bin4@3;
iy sy | cont1@0 cont2@1 cont3@2 cont4@3;
su@0; iu WITH sy@0;
*************************************************************************
Looking forward to your reply.

Respectfully,

UseMplus
 Bengt O. Muthen posted on Tuesday, December 12, 2006 - 10:47 pm
Two-part growth modeling describes growth in two parts: a binary part and a non-zero (typically continuous) part.

You can do two-part modeling where the non-zero part is treated as ordered categorical. This would allow the binary and the non-zero parts of the model to have different covariate effects. With floor effects, the lowest category would be scored missing in the non-zero part.

This cannot, however, be done in a single DATA TWOPART run because the CONTINUOUS = part is needed. You would have to first do a regular DATA TWOPART run and save the data. Then run these data with a regular input for parallel process growth modeling with a binary and a ordered categorical outcome.

Mixture modeling can be added to this, either with one latent class variable or one for each process, possibly correlating the two latent class variables.
 Mplususer  posted on Wednesday, December 13, 2006 - 2:25 pm
Thanks! If you don't mind, I would like to make sure if I truely understand you, as I am a beginner with this modeling.
(1) By "first do a regular DATA TWOPART run and save the data": Do you mean dividing my data into the binary part and the non-zero (ordinal) part. If so, I think I have done with it. If not, would you correct me?

(2) By parallel process GMM, you mean I can refer to Example 6.13 in User's Guide but treat my data as with a binary and a ordered categorical outcome and I can use regular mixture statements (e.g., Classes = (2), etc.) for these two parallell processes just as you suggested ("with one latent class variabel...two latent class variables"), right?

Thanks
UseMplus
 Linda K. Muthen posted on Wednesday, December 13, 2006 - 5:11 pm
1. Just be sure you have created the variables correctly. You can see the rules in DATA TWOPART.

2. Sounds right.
 Mplususer  posted on Wednesday, December 13, 2006 - 6:06 pm
Thanks.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: