@SPoore was investigating the wiki segmentation dataset and noticed issues in the counting of active administrators:
On my spreadsheet I made a column that listed the total number of admins for the wikis at the top of the sheet. All of them are showing more active admins than the total. Some by double or more so it can't be attrition.
From looking at the definition of active admin that you used, I'm wondering if you double or triple counted admins that used the different admin tools?