We have implemented a dual response CBC and recorded the responses into two categories. Top 3 box as 1 and bottom 4 as 0. Then the client asked whether there would be any difference to average utilities if we implemented weighting/calibration. So we tried doing that but the result for all the attributes and their levels were same except of the none attribute. Which had fallen drastically and seemed to be unrealistic.
What could be the reason for such a drastic fall in the utility score for ' none' ?