Have an idea?

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

Algorithm for MBC Combinations Counting analysis

I am planning to use my MBC software for simple "counting" analysis - specifically to determine what top ten combinations of menu items are selected by the largest number of respondents.

In my study, each respondent completes 1 menu task/screen which consists of 30 check/no check items and the respondent can select/check as many or as few of the 30 items as they wish (prices for each items are not involved in the study). So I believe this is 2^30=1,073,741,824 possible combinations that MBC needs to evaluate.  Can MBC comfortably handle a situation with this many combinations to evaluate? Does MBC have a limit on number of menu items that can be evaluated in this type of combinatorial counting analysis? In terms of how MBC conducts the analysis, does MBC brute-force evaluate all 1,073,741,824 combinations OR does it use some type of optimization or shortcut algorithm for its combinatorial counting analysis (and if so, is there any documentation available explaining the algorithm)?

Thank you.
asked Sep 26, 2016 by anonymous

1 Answer

0 votes
MBC certainly can count across all 30 items to report the top combinations that were selected.  It will report up to the top 1000 combinations.  However, it doesn't seem to make much sense to me to look beyond the top 10 or at most 30 combinations.  The frequencies get so thin on the counts beyond the top 30 or so combinations selected, it doesn't make a lot of sense to examine them.

Another thing to realize is this: let's imagine you have 2000 respondents and each person completes a single task.  The maximum number of unique combinations that your respondents will be able to select is 2000 (assuming nobody completes the BYO question in exactly the same way, which would be astoundingly improbable).  So, no, the software does not do a brute force count across all 1 billion possible combinations.  It only needs to count and report across the maximum nt possible combinations (where nt is the number of respondents times number of tasks in your MBC data set).
answered Sep 27, 2016 by Bryan Orme Platinum Sawtooth Software, Inc. (128,365 points)