In the output for Choice-based conjoint, there are some significant and some insignificant output utilities. When calculating attribute importance (using the difference formula), should I exclude insignificant variables? Or should the calculation include all variables regardless of statistical significance?
Thanks!
The difference formula gives a point estimate snapshot and as such is not a statistical tool. Therefore you should use all variables, even the ones with insignificant coefficients.
See example here, where there are some attributes with very low scores (and likely insignificant coefficients): http://conjoint.online/reports/
