choice-metrics.com

Michiel,

Great! I'll give that a go. One last question: when using constraints and pivoting in this way, how would I handle cases where a subject who has no copay enters their reference value as $0, or a subject who has strong dislike for their current facility rates it 0 stars? In the pivot design, these would then take on values less than 1, assuming x.piv[-25%,0,25%] and x.piv[-1,0,1] respectively. Would I create separate designs where each is fixed to 0, and change the pivot levels to x.piv[0%,25%,50%] and x.piv[0,1,2]?

Thank you so much for all the help, I really appreciate it.

Jamie

Pivot designs based on respondent input are sensitive to this kinds of issues. With a reference value of 0, relative pivots of +25% will still be 0.

I usually avoid this by moving away from pivot designs and instead generate separate fixed designs for different segments in the population. So in this case I would have generated in advance a design in my library for a segment that has a copay between 0 and say $2, and a star rating of 0 or 1, where the copay levels vary somewhere from 0 to $3 and star rating vary from 0 to 2 or something. I find generating these designs in advance easier than having to deal with specific respondent inputs within the survey instrument.

Michiel