Research talk:Productive new editor
Add topicWork Log
[edit]Archive
|
---|
|
Discussion
[edit]Sensitivity analysis
[edit]The productive edit definition assumes edits made to content namespace only. We haven't run a sensitivity analysis on the effects of lifting the main namespace constraint. DarTar (talk) 23:49, 18 March 2014 (UTC)
Productive editor vs Productive new editor
[edit]Research:Productive editor currently redirects to Research:Productive new editor. When we tackle the standardization and analysis of community metrics we should resolve that redirect as the two concept measure very different things. DarTar (talk) 00:04, 19 March 2014 (UTC)
Q from the analytics developers
[edit]Kevin asked:
Which implementation should be considered the authoritative example? The SQL one on the metric definition page or the Python one? Can you update the wiki accordingly?
- I don't know which SQL version you are referring to. There's no SQL demo on this metric page. --Halfak (WMF) (talk) 00:58, 17 April 2014 (UTC)
- Christian looked at the Specification section (in the right bar) and noticed it did not mention start_date or end_date, yet these were expected. This point is not so critical since Productive new editor(n, t) depends on New editor(n,t) which depends on Newly registered user which takes start and end dates as input. KLeduc (WMF) (talk) 21:21, 24 April 2014 (UTC)
- None of the specification mention a start_date or an end_date. Presumably, one can select users from a set of IDs (e.g. a campaign). Similarly, newly registered user does not take a start and end date as you specify. --Halfak (WMF) (talk) 23:12, 24 April 2014 (UTC)
Another issue: The python code relies on reverts which appears to be a custom library. The developers will also need to port this. KLeduc (WMF) (talk) 21:23, 24 April 2014 (UTC)
- Just import the library. You can install it with
pip install mediawiki-utilities
. --Halfak (WMF) (talk) 23:12, 24 April 2014 (UTC) - Forgot to ping KLeduc (WMF). --Halfak (WMF) (talk) 23:12, 24 April 2014 (UTC)
- I think we probably want to port the library to be able to use it with sqlalchemy, rather than use it as is (it builds selects on the fly) so it is consisten with the rest of the code.
- Regardless, the library makes asumptions when selecting users that are not present in the description of the metric
- (need to have a bit more time to add more detail here) 79.150.251.63 16:18, 25 April 2014 (UTC)