Subscribe Bookmark RSS Feed

Does anyone know how to apply prohibitions of level combinations using the Choice Experiment designer?

jasha_bowe

Community Trekker

Joined:

Apr 26, 2015

Hi,

I've seen one discussion here asking a similar question though I don't think the issue was resolved. I have a very simple question.

How, using the choice design module, because I am creating a design for a discrete choice experiment, can I disallow certain combinations of attribute levels.

I know it's possible to use the custom design module and use the syntax dialogue to disallow level combinations, but the designs it produces aren't proving useful for choice modeling (I run simulations on them with random binary data and get no utility coefficients). So I need to know how I can apply prohibitions in the choice design module.

I have read the advice about using prior means and variance to achieve this but they don't seem to actually answer the question. It doesn't seem like applying combination restrictions using the choice DOE module should be particularly difficult as its a common thing to want to do. ANY ONE got any tips... would be greatly appreciated.

Many thanks        

2 REPLIES
n_ryazantsev

Community Trekker

Joined:

Aug 17, 2014

This is really good question. It seems that no way to do it with JMP. I end up with SAS with #mkt family macros but this way requires much more job and I still  working on it. Please let me know if you found something.

jasha_bowe

Community Trekker

Joined:

Apr 26, 2015

Hi, I have found the same. There is no way of using custom designs with the choice DOE platform as there is no script editor or similar. I have ended up using R and a plugin which creates optimized designs in the same way that JMP does, but it simply allows you to apply level prohibitions.

The R package is called 'choiceDes' its quite straight forward which is good. It simply restricts the underlying candidate array so that any rows which include the desired prohibitions are removed from consideration in the final choice design.

Have a look at that.

Jasha