Anonymous posted on Saturday, January 08, 2005 - 4:05 pm
I am doing a latent mixture model with 4 classes, defined by 3 measures. In order to look at the differences between classes, I need to change the reference group (class 4). I have tried to enter in the estimates after each class statement, but it seems to be a trial and error process. In fact it seems as if what you put under class 1 determines the order of the rest of the classes. So, I am unsure how to make class 4 the class I want it to be. Any suggesstions would be helpful.
bmuthen posted on Saturday, January 08, 2005 - 4:11 pm
The last class often has the largest or smallest thresholds - corresponding to small or large probabilities of item endorsement. So giving high or low starting values for the class 4 thresholds often works.
Andrea Mata posted on Wednesday, April 16, 2008 - 10:33 pm
I am running a four class LCGA model with 9 covariates (6 dichotomous, 3 continous), but I want to change the reference class to a specific latent class instead of the last class. I have written this syntax command: %c#3% underneath the c on "list of all of the covariates" command. However, I do not think it changed the reference group to be class number 3. And if it did, how would I know if it did? Is this the correct syntax? I guess my question is: how do I change the reference group in a conditional LCGA model?