Have an idea?

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

How to hide attributes in an ACBC?

Dear all,

I am seeking advice on the following matter: How to „hide“ attributes in an ACBC?

I am working with the following conjoint design (attributes and levels):

Brand/model: A, B, C
Engine: Diesel, Petrol, Electric
Electric range: 300, 400, 500

Price: Summed pricing attribute

While all brands/models shall be “offered”/can be combined with all engine types in the ACBC, the attribute electric range is of course only applicable once the engine is electric, thus shall be “hidden”/not displayed when showing concepts with a Diesel or Petrol engine.

How do I set this up and what are the resulting implications on my results? I.e. do I still get sufficient information on my electric range attribute? How do I treat this in my simulation model, that is will Diesel or Petrol concepts in my simulation be weaker since they do not have an electric range?

Thank you very much in advance for your help!
asked Mar 29 by SeBo (300 points)

1 Answer

0 votes
This is called an "alternative-specific" ACBC design.  That is, the attribute for electric range only applies to the electric engine type.  To read more about how to program alternative-specific ACBC designs, see this section in the Lighthouse Studio Help:  http://www.sawtoothsoftware.com/help/lighthouse-studio/manual/index.html?alternative-specific_designs.html
answered Mar 29 by Bryan Orme Platinum Sawtooth Software, Inc. (189,440 points)
Dear Bryan,

thank you very much for the information! I studied the instructions found in the Lighthouse Studio Manual and "programmed" the alternative-specific ACBC design accordingly - it works perfectly fine!

Still I have a couple of questions to which I could not find concrete answers:

1) Using this type of design, what will be the overall effect on utility values and attribute importances?

2) More importantly, what will be the effect on my simulation model?
For example: Will non-electric cars have a lower utility and thus lower purchase probability, ceteris paribus, just because they do not have an electric range by design? Do I have to adjust those utilities manually or is this already reflected in the utility calculation (e.g. using the HB)?

3) Are there any other things I need to keep in mind to make this design work? E.g. minimum require sample size, ...

Thank you very much in advance for your help!
Glad you've figured out the alt-spec designs.  Once you learn these are available, you'll find more opportunities where they are needed & useful.  The model fits the utilities to fit the choices in the maximum likelihood fashion.  Thus, nothing needs to be done in the simulator to make this work (other than specify "not applicable" if the attribute does not apply to the product concept definition).

Importances are wonky with alt-spec design, because the importance calculation just treats each attribute as if it was independent and that the attributes should have importances summing to 100%.  I don't much care for attribute importance reporting anyway, so not a big loss there.
Thank you very much for your help and the provided explanations!
Dear Bryan,

as I have finished my research now, extracted and cleaned the data and prepared all inputs for my market simulation tool (self-built Excel based tool), I am wondering how to treat the utilities of my Electric Range attribute.

Let’s say I am constructing vehicles/models in my simulator and I have both vehicles with an electric engine (here the Electric Range applies) and some with a “normal”/non-electric engine (here the Electric Range does not apply).

What are the effects of this on the total utility of the two different kinds of models? Can you confirm that the following example calculation would be correct given the following utilities?

A = 0
B = 25
C = 50

Petrol = 0
Diesel = 25
Electric = 50

Manual = 0
Automatic = 10

Electric range:
100 miles = 0
200 miles = 15
300 miles = 30

50,000$ = 0
75,000$ = -25
100,000$ = -50

Total utility of an electric engine car:
Brand: B = 25
Engine: Electric = 50
Transmission: Automatic = 10
Electric range: 300 miles = 30
Price: 100.000$ = -50
Total utility = 65

Total utility of a non-electric engine car:
Brand: B = 25
Engine: Diesel = 25
Transmission: Automatic = 10
Electric range: Not applicable, hence +0

(so here we simply add 0 to the total utility and do not extrapolate based on the levels of the Electric Range attribute to find the Conjoint utility which theoretically corresponds to an Electric Range of "0", right?)

Price: 100,000$ = -50
Total utility = 10

Thank you very much for your support!
If you coded your utility estimation design matrix (X Matrix) properly, such that when a gas vehicle was displayed you coded the electric range attribute levels all as 0s, then this is the right way to skip the electric range attribute when it doesn't apply (you add zero utility for that non-applicable attribute).

If you used our software to do your alternative-specific design and utility estimation, then that's what our software would have done.

I'm assuming you know to use raw logit-scaled utilities when building your simulator rather than zero-centered diffs (normalized) utilities.  The example you show uses utilities scaled in a way similar to zero-centered diffs...but I suspect you did that just for illustration.