Choose Language Hide Translation Bar
Highlighted
Eva_Boonaert
New Contributor

latent class analysis for consumer research

Hello

 

I want to perform a latent class analysis on my mixture-amount experiment in JMP. Under the tab 'analyze', 'consumer reasearch', 'choice' I know how to perform a conditional logit model (via the Firth Bias-adjusted estimates) and a mixed logit model (via the Hierchical Bayes estimation) but I was wondering if performing a latent class analysis was also possible for choice experiments with JMP.

 

I tried already the tab 'analyze', 'clustering', 'latent class analysis' but there I cannot specify the specific mixture model and I cannot state that it was a discrete choice experiment​. 

 

Is there a solution within JMP to perform a latent class analysis for a discrete choice experiment?

 

Thank you very much in advance

 

Kind regards

 

Eva 

0 Kudos
3 REPLIES 3

Re: latent class analysis for consumer research

Latent class analysis in JMP is a kind of un-supervised learning method. The model yields classes that can then be used as responses or predictors. I would use the classes as levels of a response versus the mixture amounts in your case.

Learn it once, use it forever!
0 Kudos
Eva_Boonaert
New Contributor

Re: latent class analysis for consumer research

Thank you for your answer! However, I don't really understand what you mean. I included my dataset below.

 

Do you mean that I first need to perform a latent class analysis (under 'analyze', 'clustering') with the variables 'Chosen', 'Choice card' and 'Block' in my data as categorical response variables (and 'Respondent' as ID)? Then, I can save the probability for each respondent for each choice set to belong to a certain class. But how can I then use the classes as levels of a response versus the mixture components via 'analyze', 'consumer research', 'choice'?

 

Thank you very much in advance!

 

Kind regards

 

Eva

0 Kudos

Re: latent class analysis for consumer research

My apologies. I was only trying to explain LCA. No, I do not think that it would help you in this case. The choice experiments and modeling are pretty well defined. I think that there might be loss of information if you converted your responses to classes first.

Learn it once, use it forever!
0 Kudos