- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
To analyze a choice based conjoint (discrete choice) experiment I've used the phreg procedure as explained by Warren Kuhfeld in his book "Marketing Research in SAS".
One of our partners in a recent study used hierachical Bayes regression and sent us a dataset with individual part-worth utilities. The dataset showed us that there was one level of an attribute that was 'loved' by half of the respondents but the other half 'hated' the level. From the results I get from phreg I conclude that the parth-worth utility for this level is very low, which is actually a wrong conclusion given the individual part-worth utilities.
With the bchoice procedure I was able to estimate the part-worth utilities with hierarchical Bayes regression. The results are similar to the phreg procedure, but I did not found a way to generate individual part-worth utilities.
Is there a way to generate individual part-worth utilities with SAS?
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Yes, you can fit a hierarchical Bayesian model in BCHOICE to obtain individual part worth utilities, by adding the RANDOM statement.
You can find the newest documentation for PROC BCHOICE http://support.sas.com/documentation/onlinedoc/stat/142/bchoice.pdf
on SAS/STAT 14.2, where we have added MaxDiff and allocation types of choice model into the BCHOICE procedure.
There are examples provided in the documentation, in both the ‘Getting Started’ and ‘Examples’ sections. If you want, I can send you the SAS code for all the examples.
In Example 27.4, there is outpost=Postsamp specified in the proc level statement right after data=Trashcan, which requests to output a new SAS data set containing the posterior draws for all the random-effects including the subject-level random effects (individual part worths).
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Yes, you can fit a hierarchical Bayesian model in BCHOICE to obtain individual part worth utilities, by adding the RANDOM statement.
You can find the newest documentation for PROC BCHOICE http://support.sas.com/documentation/onlinedoc/stat/142/bchoice.pdf
on SAS/STAT 14.2, where we have added MaxDiff and allocation types of choice model into the BCHOICE procedure.
There are examples provided in the documentation, in both the ‘Getting Started’ and ‘Examples’ sections. If you want, I can send you the SAS code for all the examples.
In Example 27.4, there is outpost=Postsamp specified in the proc level statement right after data=Trashcan, which requests to output a new SAS data set containing the posterior draws for all the random-effects including the subject-level random effects (individual part worths).