Have an idea?

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

Including additional terms in the aggregate logit analysis


I wanted to ask if there is a method to include dummy and numeric terms to the aggregate logit analysis based on the specifications of the scenarios being evaluated in CBC.

For example, a dummy term that indicates if the worst level of each dimension is present (Wst) & a numeric term that counts the number of times these worst levels are presents (Wstn). And after running the aggregate logit for the CBC data we can get coefficients for these terms as we got coefficients for most of the levels relative to a certain level in each attribute.

asked May 6, 2020 by AMYN Bronze (3,000 points)

1 Answer

+1 vote
Best answer
I do this all the time, but it's not easily done in Lighthouse Studio.  For this I would export a CSV file (usually the dual CSV) and then add the additional variables  of interest in Excel.  Then save your work as a CSV file, import that into the standalone CBC/HB program and run the analysis.  Be careful to set your variable types to reflect what they are (e.g. partworth for categorical variables, linear or user-specified for others).
answered May 6, 2020 by Keith Chrzan Platinum Sawtooth Software, Inc. (105,750 points)
selected May 8, 2020 by AMYN
I can follow what you are saying and exported the dual .csv files. One file is for the design & one for the responses.
The term I want to add is a dummy variable that is one if level 4 or 5 is present in the scenario and 0 otherwise.
I am not sure how I can add this interaction term and to which file I should add it? ( I think I should add it to the design file a new column??)
Hi, again Keith,

the advice you gave me worked but after a weird trick from the program. If I saved the changed myself, it is not accepted & nothing is modified in the attributes, i.e. the new attribute is not accounted for. The new changes HAVE TO BE AUTOMATICALLY SAVED by the program after the error warning for them to be effective.

I hope Nothing will go wrong later no.