Have an idea?

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

Number of alternate specific attributes

Hello Folks,

I have a question on number of alternate specific attributes I can include in my conjoint design.

I have a conjoint design where we have 25 products and each has its own price bucket. One way to create such a design is to create 25 price attributes and prohibiting them with non tagged products. If I go by this design, how my overall estimation is going to be impacted. Any case example linked help will certainly help.

Thanks in advance!!
asked Apr 5, 2018 by anonymous

1 Answer

0 votes
There are a couple of ways of doing this.  One is to make 25 alternative-specific price variables, as you note.  If all the price attributes have the same number of levels, you could also do this using conditional pricing table (https://www.sawtoothsoftware.com/help/lighthouse-studio/manual/hid_web_cbc_cp.html).
answered Apr 5, 2018 by Keith Chrzan Platinum Sawtooth Software, Inc. (85,325 points)
Thanks for your reply Keith. On the first approach, is there any negative impact on the analysis if we have so many alternate specific attributes. And can I tackle 70 products in a similar way. I have been using the conditional price method from a long time, but the article still gave me few insightful points.
Not really; I've done models with scores of alternative-specific effects, including one model with over 70 of them in just the past week.   The negative impact, of course, is on run-time.  Also depending on how many brands you're showing at once, you might not have enough observations of each alternative-specific price point to support HB analysis, so you might want to run this as an aggregate MNL or maybe do some simplification (e.g. treat prices as linear or log functions rather than as categorical part-worth variables).
Thanks for quick reply. Thanks!