choice-metrics.com

Hi everybody!

I have some questions about the customization of attributes. The idea is the following: I would like to 'individualize' some of the attributes with personal information from the first part of the survey. To do that I used percentage values as attribute levels (for 2 attributes) in the design. Before the choice scenarios are shown to the respondents these percentage values are multiplied with an individual-specific number from the first part of the survey. Since percentage values do not reflect the true values shown to the respondents I was wondering whether this would pose a problem for the efficiency of the design? Is there a way to deal with such situations?
Also, I'm not sure how priors can be interpreted in this case? As related studies have used absolute values (not percentage values) it is unclear to me whether I can use their estimates as priors.

Thanks a lot in advance!
tibor

What you are doing is a so-called pivot or reference design, as covered in the manual in Ngene. The individual attribute levels serve as a reference, and for other alternatives the attribute levels are pivoted around this reference using absolute or relative (percentage) differences. So what you seem to have done seems fine with me. What Ngene does with relative differences (percentages) is, that in evaluating the efficiency of a design, it automatically converts the percentages into true levels shown to the respondent, such that the priors make sense as well. So the reported efficiency is the correct efficiency in estimation.