Multiple grp 2 part growth equality c...
Message/Author
 Jennifer Rose posted on Monday, April 07, 2008 - 9:19 am
Can you tell me how to impose equality constraints on the mean growth parameters in a 2 part growth model? I'm using the KNOWNCLASS command to specify my 3 groups. An unconstrained model shows that the slopes for both the binary and continuous parts of the outcome variable vary across groups, but I would like to be able to test whether these differences are significant. Thanks.
 Linda K. Muthen posted on Monday, April 07, 2008 - 2:46 pm
Equality constraints are imposed by placing the same number in parentheses behind two or more parameters. See the user's guide for more information.

I do not think testing the equality of these two means would make sense given that one growth model is based on logistic regression and one linear regression.
 Jennifer Rose posted on Tuesday, April 08, 2008 - 11:15 am
Hi Linda,

Thanks for your response. I wasn't clear - what I want to do is test equality of the slopes for the continuous part across groups, and then for the binary part across groups. I don't want compare the 2 slopes to each other. I have checked the manual, but can't seem to figure out how to impose the cross group equality constraints in this setting. Can you recommend an example I can look at for imposing across group equality constraints on su and sy as defined in the code below? Thanks.

classes= tx(3);
knownclass=tx (group=1 group=2 group=3);

data twopart:
names = lbinge1-lbinge6;
binary = ubinge1-ubinge6;
continuous = ybinge1-ybinge6;

model:
%overall%
iu su | ubinge1@0 ubinge2@.25 ubinge3@.5 ubinge4@1 ubinge5@1.5 ubinge6@2;
iy sy | ybinge1@0 ybinge2@.25 ybinge3@.5 ybinge4@1 ybinge5@1.5 ybinge6@2;
su@0;
iu with su@0;
iu with sy@0;
su with iy-sy@0;
 Linda K. Muthen posted on Tuesday, April 08, 2008 - 12:45 pm
The means of the slope growth factors are referred to in brackets. Equality constraints are specified by placing the same number in parenthesis after the parameters that are held equal.

[su] (1);
[sy] (2);

If you continue to have problems, send your input, data, output, and license number to support@statmodel.com.