Choice Based Conjoint -MNL

[Fabio5]

Is it possible to run a Choice Based Conjoint analysis with SPSS?

Otherwise if not possible would it be the same running a Multinomial Logistic regression?
Does MLN provide the same info of Conjoint analyses (attributes importance, simulation of preference etc)

Thanks in advance,

Fabio

 

[Injektilo]

It's possible, but quite difficult. If you have any access to Sawtooth software, I'd recommend it over using SPSS.

There are three ways to run analysis on a CBC:
1. Hierarchical Bayes MNL Regression. This would be the most preferable option, but I'd have no idea how to run this in SPSS. It is possible to run it in R using the bayesm package, but it's not an easy task. Check out the following blog posts for some help:
http://joelcadwell.blogspot.ca/2013/03/lets-do-some-hierarchical-bayes-choice.html
http://joelcadwell.blogspot.ca/2014/11/lets-do-some-more-hierarchical-bayes.html

2. Latent Class Analysis. This is particularly useful to find segments in the sample based on their preferences as measured in the CBC exercise.

3. Multinomial Logit Regression. While easiest, it is the least desired option because it is open to the independence from irrelevant alternatives (IIA) issue, which is very problematic if you are looking to simulate market scenarios with options appearing and disappearing.

I would suggest you reading up on the literature on running CBC analysis on Sawtooth's website:
http://www.sawtoothsoftware.com/support/technical-papers#cbc-related-papers

 

[Fabio5]

Thank you very much for your reply.

Regarding your 3rd option : MNL and the issue of IIA. If I do a Choice based conjoint and each time I ask my respondents to choose among the same choice set of brands (hence I am simplying changing the price and other attributes but the same 4 brands are always present in the choice set), then will my MNL suffer from this IIA?
Furthermore, How can I create an ortogonal design for my CBC? meaning, how do I know how to vary attributes levels between the tasks?

Thank you,

Fabio

 

[Injektilo]

IIA is only a problem if you intend at the analytical stage to create scenarios in which a product will appear in the market or disappear from the market. It's also known as the red bus/blue bus problem. A description of the problem can be found here: http://en.wikipedia.org/wiki/Independence_of_irrelevant_alternatives#In_econometrics . If you always intend your scenarios to include only the 4 products you are testing, then don't worry about it.

In regards to setting up an experimental design, I use Sawtooth software to create them for me in a pinch. If you don't have the software available, I would check out the R package "Conjoint": http://cran.r-project.org/web/packages/conjoint/conjoint.pdf

See page 4 for an example of how to set up a design using that package.