Have an idea?

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

Sum Allocation Analysis in CBC/HB

Hi Team,

I'm running analysis for sum allocation CBC in CBC/HB tool and would like to understand "Total task weight for constant sum data" in Data Coding tab and how best way to use it (Do we have any paper which I can refer).

Below is the background of the study:
The Design consist on 3 Attributes i.e.
Attribute 1 with 7 levels,
Attribute 2 with 3 levels and
Attribute 3 with 7 levels.
Also, design is alternative Specific Design i.e. Attributes 1 and level 1 only seen with Attribute 2 and Attribute 3 levels and remaining level of Attribute 1 i.e. 2 to 7 will have blank/zero for level for Attributes 2 and 3.
asked Sep 14 by Milan Sanathra (490 points)

2 Answers

0 votes
The task weight is the setting to allow you to basically determine how much weight to give your allocations.  For example, if you asked people of their next 10 shopping trips, how many items they would purchase, you could choose to set the task weight as 10.  This would treat each allocation as a separate decision as far as the utility modeling is concerned.  A weight of 5 would half the weight and not treat each allocation as a separate choice, and a weight of 1 would treat it like a normal discrete choice.  You can think of it like repeating the choice task in the data set.

It's hard to know exactly what the weight should be.  The software defaults to 5, but it's kind of up to you to decide how much do you think people are just putting allocations together, versus thinking about each one independently.
answered Sep 15 by Brian McEwan Gold Sawtooth Software, Inc. (47,900 points)
0 votes
Appendix F in the CBC/HB Manual, which you can download in PDF form from: https://sawtoothsoftware.com/resources/manuals

has the most complete information regarding how constant sum data are treated in CBC/HB utility estimation.  It also provides an explanation regarding "task weight".
answered Sep 15 by Bryan Orme Platinum Sawtooth Software, Inc. (177,015 points)