Laura Tang posted on Wednesday, April 22, 2009 - 3:06 pm
We are trying to use Mplus to do some post-estimation for our latent class model (LCM) with a binary choice situation (intention to use store versus internet for a future purchase). Our questions are: (1) How to compute elasticities for a LCM with a binary choice? What command we should use? (2) How to compute the elasticities for dummy variables? Will that be computed automatically as well as other continuous variables?
I tried to search the related key words in user's guide but couldn't find the answer.
Laura Tang posted on Wednesday, April 22, 2009 - 3:09 pm
In addition, our sample is a choice-based sample, thus we need to include the weight variable into the estimation, how to do that?
(1) Perhaps you are referring to the regression coefficients when regressing the binary choice variable on latent class (i.e. a set of latent class dummy variables). This is accomplished by default when including the binary choice variable on the USEV list in that the threshold (negative logit intercept) and therefore the probability of the binary variable varies across the classes as the default. So the result is a set of logits/probabilities, not regression coefficients, but you can translate those results into reg. coefficients.
(2) It's the coefficient for a change from 0 to 1 in the dummy, right?
Choice-based sample analysis would need to use the weight option for logistic regression in line with the choice-based literature.