User:Mike Christie/FACstats/Data description

This page describes the FACstats data.

Data
There are three main tables. As of now (December 2020) they contain data from October 2006 through November 2020.

FACs


 * Article title
 * Archive number
 * Nomination date
 * Outcome (A for archive, P for promotion)
 * Outcome date

Nominators


 * Article title
 * Archive number
 * Nominator

Reviews


 * Article title
 * Archive number
 * Reviewer
 * Review type (blank, I, S, or A)
 * Declaration (S, OS, SO, SS, O, or blank)

What is included
This is a list of how I decided what to enter.


 * Supports and opposes were only entered if bolded, unless it was absolutely clear what the intent was. A withdraw is treated as an oppose, but refer to peer review is not.  A procedural close comment is treated as an oppose.  Adding "conditional", "weak", or "strong" makes no difference. Humorous opposes are ignored, struck or not.
 * If the closing coordinator makes substantive comments then they are credited them with a review, but not if they make a minor comment (e.g. "I see some duplinks and periods on captions, please check throughout"). This was hardest with Sandy's closings as she often posted a couple of comments as she was considering closing.  At around three or more bullet points I treat this as a review.  If a coordinator comments and does not close the FAC they are treated like any other commenter.
 * Nominators' declarations of support are not included. This includes conominators, even if they are added as conoms after they have reviewed.
 * Comments that do not help the FAC in any way are not recorded; this is interpreted narrowly. E.g. humorous comments are excluded and so are posts such as "Did you ask major contributors before nominating?", but "You should have asked major contributors before nominating" is recorded because it informs other reviewers.
 * If a review only addresses image issues, or only addresses source issues, but is not marked as an image or source review, it is a judgement call as to whether to count it as an I or S review type. Other reviewers who respond to conversation under an I or S review are usually not marked as I or S reviews.
 * A review that covers only one aspect of a source review is still counted as a source review -- e.g. a source formatting review and a spotcheck. If these are done by two reviewers it counts as two reviews; if they're done by the same reviewer it will only count as one S review.
 * If a reviewer does a source review and an image review and a content review they are recorded as having done three reviews.
 * Supports and opposes are only marked on the image and source review if it's clear that's the reviewer's intent, otherwise the declaration is placed on the content review. It is possible to record two (or three) opposes for one reviewer, but only if it's absolutely clear that the reviewer meant to oppose on e.g. both sourcing and images.
 * For a restarted nomination, the archive number is given .1 and .2 suffixes, so that each version of the review can be referred to separately. See Featured article candidates/Pepper v Hart/archive3, which was restarted, and recorded as archives 3.1 and 3.2; there is some material from the first review that was left by the coordinator in the second review.  These comments are not tabulated again in the second review, but if those commenters comment again they would be counted.
 * Conominators are included only if declared, and if their name is listed at the top of the FAC. This is fine if done partway through but see for example Featured article candidates/Paul Kelly (musician)/archive2, where I did not include Shaidar cuebiyar as a nominator, since they were not listed at the top.  This is a rare situation.
 * If a nomination is transcluded late, record the start date as the date on which it was transcluded, not the date on the nomination statement. In some cases the reviews that predate transclusion were moved to the talk page; I counted these.
 * User names are particularly complicated, since several users have changed their names, sometimes repeatedly. I used the name that shows on the edit, meaning the current user name of that user, except for vanished users, for whom I used the name in use at the time they vanished -- typically this is also the name visible on the FAC.  I have kept a record of all name changes which I use to look for mistakes in the data, and I will make that list public at some point, but it does not include vanished users.
 * If two users are known to be sockpuppets (e.g. DCGeist and DocKino) I made no attempt to connect them.
 * In a few cases where there are obvious trustworthy connections between usernames I substituted them -- Giano II is Giano; Lingzhi2 is Lingzhi.
 * If a declaration is struck by someone other than the reviewer (almost always because the reviewer is a sock) I still count it as struck.
 * Source reviews began in earnest after 31 October 2010, because of a mainpage copyvio, but there are still plenty of explicit source reviews prior to that date, e.g. from Brianboulton, Ealdgyth, and Fifelfoo.

Caveats

 * Quite a few editors had a habit at times of doing a specific verification of some kinds across multiple articles. For example, Graham Colm did Copyscape checks; GermanJoe did link checks; Cryptic C62 did lead reviews; Lightmouse did units reviews; PresN did EL/DAB checks, as did Ucucha and PL290.  These cause those reviewers numbers to be higher than for reviewers who do more in-depth reviews of each article.
 * A one-line "Support" with no commentary counts as one review; so does a 5,000 word detailed critique. This means it is not possible to tell the value of a review from this data.  The most that can be said is that there is a correlation between a high review score and value to the FAC process, but it is not the case that of two reviewers, the one with the higher review count is definitely the more useful reviewer.
 * If someone deletes their support or oppose instead of striking it, it is unlikely to be recorded correctly.

Unusual situations

 * The nominator for Galapagos tortoise abandoned the FAC, so Sandy added several of the reviewers to the nominator list, leaving only one support from someone who was not a nominator.
 * As a curiosity, the last date on which a nominator felt it necessary to declare "self-nom" was 16 December 2010, here.