In our standard conjoint utility run we run individual attribute importance on a respondent by respondent basis as part of our high level reporting. My understanding is that this is based on the variance (max-min) of ZC values, which are then rescaled so that they sum to 100, correct?
Recently we've transitioned from primarily using Top Choice methodology to Share of Preference, which has worked quite well for us. Does Sawtooth have a position on how to calculate individual attribute importance levels based on the Share of Preference methodology? Should we be taking the max EXP value - min EXP value for each individual attribute to determine this range, or should we perform the EXP transformation on the range itself? Or in other words should do
=EXP(max [Lvl1 through Lvln] - min [Lvl1 through Lvln])
=max [EXP(Lvl1) through EXP(Lvln)] - min [EXP(Lvl1) through EXP(Lvln)]
Are am I wildly off base about these calculations, and need to be doing something radically different?