Have an idea?

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

Converting CBC data to Stata


I'm trying to export CBC data of a survey to Stata. However, the data is not ordered or structured in a way that is acceptable in Stata or to run models in Stata, such as conditional logit model or mixed logit model.

I need the data in a different way for Stata. Let's say if I have 9 choice sets (tables) with 3 options per choice set, then I would need 27 rows per participant. I will need details about the attribute levels that were presented to each respondent. How do I get the data in this structure with corresponding respondent socio-demographic characteristics?

asked Mar 3, 2018 by Anonymous

1 Answer

0 votes
Assuming you are using CBC within either Lighthouse Studio (or SSI Web), open your project that includes your CBC exercise.  Go to File + Data Management...

Click, AddJob... , and then click Add again.  Select the file type "CBC (*.csv , .cho/chs, .att)".

On the File Format option, click "Sincle Format CSV".  This will put all the attribute level information shown to respondents in each CBC question along with the respondents' answers within the same .CSV file.

Click OK.  Then, Click Export Jobs.  A .CSV file is written to a file folder within your project.  

This .CSV file has your data arranged as 27 rows per respondent (assuming you have 9 choice sets with 3 options per choice set).  

To get respondent-based demographic characteristics also in this file, you will need to export your respondent characteristics separately using the Data Management area.  Then, you'll need to do some extra data processing to put respondent characteristics in the CBC .CSV data file that has 27 rows per respondent.  I don't know how Stata is supposed to accept those data, so you'll need to refer to their documentation.
answered Mar 4, 2018 by Bryan Orme Platinum Sawtooth Software, Inc. (160,785 points)
Thank you, I see that I would have to do a lot of work in order to match respondent socio-demographics to attribute levels and choice sets. It is doable, but time consuming. It would be useful if Sawtooth Software give us this data in this structure (number of rows = choice sets * options per choice set, and corresponding socio-demographics), so we can use it in other statistic software that we are familiar with.

Best regards