Profile: How do I interpret Dimension Level Summary Statistics?

It is possible to view Dimension Level Summary Statistics on a Profile report. These statistics rank the overall power of the dimension as a predictor.

Right clicking on a column header allows you to add the statistical measures Phi and Cramer’s V.

Dimension Column Summary

Mean Index - weighted mean of (equivalent positive) Index across significant categories
Min Index - min of Index across significant categories
Max Index - max of Index across significant categories
Mean PWE - weighted mean of the absolute PWE scores across significant categories
Mean Z Score - weighted mean of the absolute Z scores across significant categories
Phi & Cramer’s V - standard statistical measures

The Mean Index, Mean PWE and Mean Z Score are all weighted by the proportion of analysis records in each category.

Categories that are not significant (have not reached the threshold Z-Score) are excluded.

The weighting reflects the distribution across the categories and prevents a large index that only applies to a small number of records dominating the summary result.

To calculate the Mean Index for each category, we first take the equivalent positive Index.

Combining directly positive (100-infinity) and negative (100-0) indexes together makes no sense as they use different log scales. Therefore, (100/Index)*100 is used to convert from a negative to a positive index.

Example: Calculating Mean Index

The example below relates to the Gender Dimension - see screenshot above - and explains how the Mean Index of 193.31 was derived.

Female: ZdExp > 3 significant --> category index 52.10 convert to positive index (100/52.10)*100 = 191.92 weight by 8674/25175 --> 66.13
Male: ZdExp > 3 significant --> category index 195.39 weight by 16101/25175 --> 124.97
Unknown: ZdExp > 3 significant --> category index 139.32 weight by 400/25175 --> 2.21
Summing across categories 66.13 + 124.97 + 2.21 --> 193.31