User:Amgine/Wikidata random sample test 2

This is a simple check to examine recent user contributions on Wikidata vs WMF global contributions. My hypothesis is at least one of these 10 contributors will have reduced their non-Wikidata contributions, or xyr most recent edits will solely be to Wikidata, which is a reasonable proxy measure of contributor poaching.

This is variant on a theme; LauraHale suggested a much stronger study using a better methodology (and easier tools.)

Methodology
Using a snapshot of recent changes on Wikidata to identify a random sample of human contributors (Grabbed 500 recent entries, selecting the first unique username after rows 0, 50, 100, 150, 200, 250, 300, 350, 400, and 450.) Used Global account manager (GAM) to identify each user's total contributions to Wikidata, their 'home wiki', and the wiki they have the largest total contributions (if different.) Where an account name is not found via GAM, Global user contribution beta tool is used as a fallback. This tool does not report 'home wiki', so only the largest total contributions wiki is used.

With these three wikis identified, use X!'s Edit Counter to break out the contributor's monthly contributions on each. Simple comparisons are the only ones used due to the unreliability of the data. (The users found use their SUL accounts for automated editing, therefore simple contribution numbers are not a reliable measure of editor engagement.)

Discussion
100% of the sampled human contributors to Wikidata have previous WMF project involvement prior to editing Wikidata.

The human contributors to Wikidata make use of automated processes for editing using their SUL accounts. This makes the revision count a weak measure for user engagement with a wiki.

Half of the sample did not opt-in to directly measure their contributions over time. Additional coding would be necessary to extract this information directly from their contribution history, and I'm not willing to put in that time for this simple study. Of those for whom we do have monthly data, at least one (AmaryllisGardener) appears to have an increasing or steady trend on Wikidata, while involvement on home or highest contribution sites have steady or decreasing trends. Of those who have not opted-in, 2 (Place Clichy and Thieol) have higher averaged rates of involvement with Wikidata over their more-longstanding home wikis.

Conclusions
The evidence strongly suggests Wikidata competes with sister projects for contributors, however the measures are not effective and may not reflect contributor effort or engagement.