Have an idea?

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

Is it possible to include multiple constant alternatives in a partial profile design?


I am interested to know whether it is possible to include multiple constant alternatives when specifying a partial profile design.  For example, is it possible to modify the .CHO file to do this?

All help and advice would be greatly appreciated



The CBC software does not support or document the use of a design that is both alternative specific and partial profile.  I am now interested in the possibility of creating a partial profile design that has several constant alternatives, that remain fixed in each task.  These would be accompanied by 2 or 3 alternatives that would vary according to the attribute levels.

The SSI Web Help and the Advanced Design Module manual state that more than one constant alternative can be created by using an alternative specific design.  Simply include the constant alternatives as levels of the primary attribute, and then specify prohibitions between these constant alternative levels and all of the levels for all of the other attributes in the design.
asked Oct 31, 2012 by TJ Bronze (970 points)
retagged Oct 31, 2012 by Walter Williams

1 Answer

0 votes
Best answer
I think the answer is that this is possible, and one way to do it is as follows:

1. Generate a partial profile design with the partial profile attributes and alternatives.  
2. Export the design and manually add an additional attribute, containing one level for each partial profile alternative, and additional levels for each constant alternative.  
3. Specify the partial profile levels to be 0 for the constant alternatives.
4. Copy your CBC.ssi file and change the design to match the one you are going to import.
5. This means adding your new attribute, and changing the design to be alternative specific.
6. Test your design using the advanced test to make sure you have good standard errors.
This is what I have done, and I think the process is the right one to do.  

Clearly this approach only works when one is happy with the realism of respondents comparing the partial profile alternatives to the constant alternatives.  

If anyone has any comments to add please let me know!  

answered Nov 2, 2012 by TJ Bronze (970 points)
selected Apr 22, 2014 by TJ