

Freeing intercept and slope with bina... 

Message/Author 

Corinna posted on Wednesday, June 03, 2015  6:34 am



Dear Linda & Bengt, I want to free the variances of my intercepts and slopes for a binary outcome variable with 3 classes. If i formulate %c#1% i*; s*; an error occurs. Can you tell me why? The default for GMM with a binary outcome is that variances of intercept and slope are hold equal across classes. However, my model seems to fit better if I fix the variances of intercept and slope at 0. Could you tell me which conditions must exist that it makes more sense to fix the variances at 0 or to freely estimate them? I would be very thankful for a brief answer. 


We need to see your full output  send to support. Let BIC decide which model fits better for you. Typically people vary within classes, calling for nonzero i and s variances. 

Corinna posted on Sunday, June 07, 2015  4:02 am



Dear Linda & Bengt, I am very thankful that I can send my output to you. However, before doing this, it would be necessary that I knew how to correctly free intercept and slope with a binary outcome... If I compute the following syntax, the variances of i and s are still hold equal across classes but are not freely estimated across classes as they should be...? Consequently, my syntax seems to be incorrect. Could you tell me the right one? When I have results for that, it probably makes more sense to send my output to you... Thank you very much for your help! This is my syntax: variable: names = sex age KSty sch KS1 EKSW1 LEw KS2 KS3 KS4 rD rSC rSB rKli; usevar = KS1 KS2 KS3 KS4; missing = all(999); classes= c(2); CATEGORICAL = KS1 KS2 KS3 KS4; analysis: type = MIXTURE; STARTS = 400 100; ALGORITHM = INTEGRATION; model: %OVERALL% i sKS1@0 KS2@1 KS3@2 KS4@3; [i s]; %c#1% [i s]; %c#2% output: TECH1 TECH8 PLOT: SERIES = KS1KS4(s); TYPE = PLOT3; I also tried the following: %c#1% [i* s*]; %c#2% [i* s*]; 


%c#1% i* s*; %c#2% i* s*; 

Back to top 

