Confidence Interval for Importances

Question

Dear Bryan,

thanks for the quick reply, the importance scores are certainly needed.

If the Importances should be calculated on the individual level, than how would you recommend to calculate the confidence intervals?

With respect to the earlier described approach (mean betas), the resulting importances are very similar to the importances reported in the utility report. The level hierarchy  is also fairly homogeneous when looking at the individual PWU.

Best,
Frank

Answer 1

Typical approach is to develop confidence intervals using the Frequentist approach:

Compute a single estimate of each respondent's importance scores (using the collapsed point estimates for each respondent).  Our software does this automatically.

Take those individual-level importance scores, and using Excel, R, or similar, create the 95% confidence intervals.  That's done by taking the mean importance score +/- 1.96x the standard error.   The standard error is computed by taking the standard deviation of the importance score and dividing by the square root of the sample size.