Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Data formatting for evaluating perceptual choice experiments

This question is in regards to the recent paper written by Bryan Orme here: http://www.sawtoothsoftware.com/support/technical-papers/169-support/technical-papers/cbc-related-papers/1502-perceptual-choice-experiments-2015

I'm a little confused in regards to how the data from the perceptual choices should be coded. Exhibit 6 on page 7 of the paper shows the "availability" IVs for the items being coded 1 or -1, depending on whether it was presented along with the 'best' or 'worst' concept (and I'd assume coded 0 if the item was not available). However, Exhibit B1 in page 25 shows a screenshot of the raw data file in the MBC software, with all "availability" IVs being presented as 1=available, and 2=Not available.

Is the analysis presented at Exhibit 6 separate from the one presented at Exhibit B1 then? Would it be possible to have an explanation to reconcile or tie in the 2 exhibits?
asked May 11, 2015 by anonymous

1 Answer

0 votes
Thanks for the interest in the paper.

Exhibit 6 is the coding for estimating the determinance (importance) of each perceptual item.  We're just estimating how much agreement there is between choice for product concepts and that those products are associated with certain perceptual items.

Exhibit B1 is the coding for estimating the part-worth utilities (for the conjoint levels) as predictors of perceptual associations.  For example, how the different cities, hotel options, car rental options, and prices for a vacation package "drive" the perceptions.

In the first model, you just get a single coefficient per model, indicating how likely that perceptual item is associated with choice.

In the second model, you obtain many parameters (because we have many attributes and levels in the conjoint experiment) that indicate how the parts of the product concept (city, hotel options, car rental options, prices) drive perception with specific perceptual items.
answered May 11, 2015 by Bryan Orme Platinum Sawtooth Software, Inc. (169,815 points)
I just used the "determinance" model to decide which perceptual items to include in the heat maps.  Those perceptual items that didn't seem to be associated with choice were dropped.

The second model (actually a model per each perceptual item) was used in the Excel simulator to create the perceptual heat maps.
Good morning, all.

Bryan, regarding the design of the sentences shown inside every CBC task, was it similar to a MaxDiff experiment?

Thanks a lot for your time.