[Wikimedia-l] Thanking anonymous users

[Mark]

On 1/17/14, 3:55 AM, Erik Zachte wrote:
> Here are some charts which breakdown edits into several categories, reverts are counted separately. Of course edits is not editors, but it could be indicative of changed behavior patterns/policies. In the ongoing reassesment of metric definitions one thing discussed is whether we should count productive editors separately (I think we do), and if so on what basis (e.g. x edits per week/month which survived y days of not being reverted).
>
> http://stats.wikimedia.org/EN/PlotsPngEditHistoryAll.htm
>

Thanks, that's interesting information!

Edits-that-survived is one baseline definition of productivity, though 
I'd be interested in data with a higher threshold for productive edit, 
if there's some way to distinguish them. For example what I really want 
to know is what the trends are for article-writing activity, e.g. how 
many editors we have who regularly expand articles with new information. 
Spot-checking some articles, these edits are by far in the minority, so 
in edit counts they are typically *completely* swamped by other kinds of 
editing. For example, picking one random article with a short edit 
history, here is a manual classification of its 12 edits:
- 2 substantial content additions (original author, plus 1 subsequent 
author who added a paragraph)
- 7 bookkeeping/formatting edits (category sorting, maintenance tags, 
infobox, bot replacing a template, etc.)
- 3 copyedit edits

Those are all valuable in various ways (and I do all three kinds of 
editing myself), but a minority of the edits (2/12) have produced the 
vast majority of the article content (and all of its references). What I 
fear with the aggregate edit counts is that minor changes in bookeeping 
edits swamp even significant changes in the content edits, because the 
content edits are a relatively small percentage of the total count. So 
I'd be interested in seeing data on those categories separated out, 
though it's admittedly difficult to collect.

One mechanically countable statistic might be a count of edits that 
include a reference. Would miss references not added with a <ref> tag or 
other easily recognizable pattern, but might produce interesting trends 
nonetheless.

-Mark