Have an idea?

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

Some hint for an appropriate approach

Hey guys,

Currently I am facing a conjoint problem for which I could need some advice for an appropriate approach.

Just think of a car interior in which we distinguish different areas, e.g. the seat or the steering wheel, and so on.
For those areas we got different attributes, e.g. for the area seat attributes like color, material, existance of seat heating, and so on. Those attributes consist of different levels, of course, like it is usual for a conjoint study (Color: black, white, green; seat heating: yes, no, ...)

So in total, we got a quite large number of attributes (about 25 to 35) that we can assign to 10 different areas. The number of levels for the attributes range from 2 to 6.

Now our objective is to get importances for the attributes but also for the global areas.
I wonder what could be an appropriate approach for this study - ACA, CBC or maybe a combination of those?

In some way it would be useful to only show attributes of one area when using CBC tasks, for instance. But as we also need the comparison of the areas, we cannot only use a partial profile CBC. But those are only my first thoughts.

Many thanks in advance for any hints or ideas
asked Jan 23, 2013 by TomHenning (140 points)

Another possibility to consider is that this might well fit as a menu based choice (MBC) problem, about which you can read more in the technical paper:  https://www.sawtoothsoftware.com/download/mbcbooklet.pdf.

Yet another way I've seen people do this is to come up with a “high” “medium” and “low” version of each of your areas (interior design, audio technology, exterior).  Then a given respondent receives a design where she chooses among profiles in which the attributes and levels of one area (e.g. the interior) vary independently but the other areas move together in the high/medium/low blocks.  

It might take some digging but I believe there’s an academic article out there about this kind of design.   I do not recall this method being widely adopted as it is still pretty complex.

I believe this is an aggregate analysis that is not amenable to HB modeling.

1 Answer

+1 vote
Hi, Tom.  

Regarding your last paragraph, if you need to be able to compare the different areas, you will not be able to have some areas with their own distinct CBC exercises, because then their utilities would not be comparable.  I'm afraid you have a decision to make here about which of your design objectives are more important than others.  

With 25-35 attributes you could well do a partial profile design and it would allow you to compare utilities across areas (but it would not allow you to do so if you want to have a compartmental sort of design where some car areas have their own CBC)

answered Jan 23, 2013 by Keith Chrzan Platinum Sawtooth Software, Inc. (52,325 points)
CBC with partial profile