Have an idea?

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

Is it possible to analyze utilities for different products of one CBC?

Dear Sawtooth Team,

I have performed a CBC that includes combution engine and electric vehicles (attribute propulsion type). The CBC contains some conditional relationships, otherwise the other attributes (e.g. price) do not distinguish between the propulsion types. Would it still be possible to evaluate combution engine and electric vehicles separately, i.e. evaluating the utilities for combustion engine vehicles and evaluating the utilities for electric vehicles? This should allow to analyze the differences afterwards.

Thanks a lot!
asked Jul 30 by Annkathrin (140 points)

1 Answer

0 votes
It's possible, yes, but it depends on the details of your design.  The most straightforward way to do this is using what's called a "labeled" design.  But if you didn't do that it's still possible assuming you used our software to make the design and if your design had enough versions/blocks.  In that case you just need to include the interactions of propulsion type with all the shared (i.e. "generic") attributes (all the attributes combustion and electric vehicles share).  Personally, I'd find it easier to export the design and recode it as if it were an alternative-specific design (to estimate the propulsion-specific values of all the shared variables) but a lot of folks are comfortable doing this with interactions, and it should get you to the same place.  

If you try to estimate interactions and your model crashes, that's evidence that your design won't support getting the separate effects of the shared attributes - so hope that doesn't happen.
answered Jul 30 by Keith Chrzan Platinum Sawtooth Software, Inc. (114,400 points)
How exactly does the "labeled" design work? And how do the results look like?

I also like the idea of the alternate specific design. Does this also work when multiple attributes depend on the same primary attribute? The importances would then also be propulsion specific (despite the primary attribute), right? When plotting the part-worth utilities as a graph, there would then be two curves, for internal combustion engine vehicles and electric vehicles?

I then have another very specific follow-up question. My colleagues have previously done a very similar CBC, with the same attributes, but only related to combustion engine vehicles. Is it possible to combine the two conjoint surveys by extrapolating the results of the new survey to the results of the old survey (by comparing the results?)? I mean, using the values for combustion engine vehicles of the two surveys as auxiliary variables to adjust the results of the new survey to the values of the old survey? I really hope that my description is understandable!

Thanks a lot and have a nice weekend!
...