Have an idea?

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

Adjusting utilities from 2 CBC studies to another

I have following problem:
There are two CBC studies with different attributes but for the same product.
One attribute (costs) is included in both CBC studies and has the same attribute levels.

So I have one CBC study with A, B, C attributes plus D (costs)
and an other CBC study with E, F, G attributes plus D (costs).

The idea was to adjust the utilities of both studies to another and use the common attribute cost as basis.

I already have the part worth utilities for both groups and calculated the relative importances for each attribute and each respondant.

How can I adjust the relative importances of both CBC studies to another so that I can put them in one table?
asked Nov 5 by Shoel (140 points)

1 Answer

0 votes
I'll assume that you've already asked yourself the question about "should you combine the utilities into one table" and answered in the affirmative.  Of course there are lots of worries about doing this and some potential for overcounting the value of some attributes and undercounting the importance of others, but, again, we'll assume that you've considered those worries and judged that they don't apply to your survey.

In this case, you could go back and rerun the CBC analysis, including all the choice sets in a single analysis.  You would go back and recode the data matrix so that you had 0 codes (for "missing") for attributes E, F and G when you were modeling choices from the first study and then 0 codes for A, B and C when you were modeling choices from your second study,  If you concatenate the two data matrices  and run them as a single model, well, there's your utilities, all in one model.

If you don't have access to the raw data, a very simple thing (not ideal, because better would be combining the data into a single model as described above) to do would be to use the cost utilities from study 1 to predict the cost utilities in study 2.  For example, if you have 5 levels of cost, then in study 1 your 5 cost utilities are x1, x2, x3, x4 and x5 and y1-y5 are your cost levels for study two.  Now run a simple linear regression analysis to predict your y variable with your x variable and voila, you have an equation that calibrates your study 1 utiliteis to fit with your study 2 utilities.   This is one low tech way of "bridging" utilities between two sub-experiments.
answered Nov 5 by Keith Chrzan Gold Sawtooth Software, Inc. (48,525 points)
...