4
$\begingroup$

StackOverflow launched its beta-Documentation page about a year or two ago. The response there seems to be very positive. I am wondering if this site would benefit from something similar and what the structure would look like.

Personally, I think it'd benefit many new visitors to this site to have the option of seeing a reference manual for a type of statistical tool (e.g. linear regression, logistic regression, neural nets).

I also think it'd be great to consolidate some of the great answers provided in some of the site's posts.

Edit:

To clarify what 'documentation' may look like to serve CV visitors, I am providing an example:

MIXED EFFECTS LOGISTIC REGRESSION (UCLA)

I would like to see what feedback is on this type of documentation.

$\endgroup$
  • 3
    $\begingroup$ "If you have told us (or privately thought) Documentation isn't working: You are correct. It isn't. Yet." The title of this post indicates just how far this product is from coming to other sites. To have Documentation on CV, one needs to have working Documentation on SO first. $\endgroup$ – user56674 Jul 14 '17 at 20:16
  • 2
    $\begingroup$ (+1) Thanks for alerting us to this! The SO Documentation tour says:-"When we reviewed traditional documentation, two things were clear: (1) It had to be based on assumptions, not need. It was usually written once, often by someone not even using the technology, so it was a guess at what to focus on. (2) It didn't prioritize good examples. People learn best when they can see things demonstrated in actual code." What would be the raison d'être for ours? Why consult it rather than one of the many textbooks or websites covering linear regression &c.? Why contribute to it rather than to Wikipedia? $\endgroup$ – Scortchi - Reinstate Monica Jul 14 '17 at 20:18
  • 1
    $\begingroup$ @Jon: You seem to have an idea in mind that you're contrasting with other reference sources but haven't shared with us. Why not detail it in an answer? $\endgroup$ – Scortchi - Reinstate Monica Jul 14 '17 at 20:55
  • 5
    $\begingroup$ 1. Documentation is about how to use a particular aspect of some language (or similar) -- explaining "How do I use A to do B?" is usually specific and clear; indeed in some languages there's arguably only one right way to do most things. Statistics is a bit different from that because much of it's not so clear cut, requiring more teaching about something than giving a recipe for it (indeed, I see focus on recipes in place of understanding as a large part of the problem we have). Given the plain differences from documenting a language, what is it you'd anticipate this producing? ...ctd $\endgroup$ – Glen_b Jul 15 '17 at 9:37
  • 3
    $\begingroup$ 2. If your concern is about expense, note that there's a plethora of free statistical textbooks. As with the more expensive ones some are good and some are terrible, but there are some good free ones. What will we be doing that wouldn't be covered by writing a textbook with a slightly different emphasis (such as on better examples)? ... 3. A big concern I have is who would do this. We already have a pretty low answer rate here -- we lack sufficient people able to write good answers to nearly half the questions we get. If regulars redirect their efforts elsewhere, what happens to that? $\endgroup$ – Glen_b Jul 15 '17 at 9:43
  • 3
    $\begingroup$ You should not read this as opposition to the idea; I mostly seek clarification of what is being proposed and some sense of the scale of effort and consequences. $\endgroup$ – Glen_b Jul 15 '17 at 11:38
  • 1
    $\begingroup$ I liked documentation on SO at first, but then came to find it not worthwhile. I don't think I've ever contributed & I never use it. I just read the real documentation or find an existing question that explains whatever the issue is on r-help or SO. Like others, I don't see the analogy to a 'documentation' on CV. $\endgroup$ – gung - Reinstate Monica Jul 16 '17 at 0:38
  • 4
    $\begingroup$ The practical question is not whether this would be a good idea, like world peace or Western civilisation, but how far it is all likely to get off the ground in any serious way. Evidence for this community is documentation on R stackoverflow.com/documentation/r/topics. If there's a large group of well-informed users of any statistical software active on SO, it's surely R users, but after a year or so it seems unlikely that -- whatever the quality of material posted so far -- documentation that matches what is already well documented in many forms can be produced in this way. $\endgroup$ – Nick Cox Jul 17 '17 at 14:25
  • 1
    $\begingroup$ More candidly put, a wild guess is that very many people don't think that trying to write new documentation from scratch is a good practical idea. That's certainly my personal opinion about Stata on SO. I don't want to disparage anyone with different ambitions on SO or on CV, but it's my impression that there aren't enough of them for this to work, there or indeed here. $\endgroup$ – Nick Cox Jul 17 '17 at 14:28
  • 2
    $\begingroup$ On the UCLA documentation (where I have some extra knowledge as having met former personnel from there and heard how they worked): that site depends rather crucially on a tight and agreed house style and liaison between a fairly small group of people. Those are precisely the details that are more elusive in a forum like CV. $\endgroup$ – Nick Cox Jul 17 '17 at 19:22
  • 2
    $\begingroup$ @Nick indeed in relation to Python precisely that claim is made (there's one right way to do it) ... and that makes it a good candidate for documented examples of particular tasks (how do I do X? In Python, you do it this way,..). Statistics is a different kind of beast I think. Guidance and principles. $\endgroup$ – Glen_b Jul 17 '17 at 21:56
  • 2
    $\begingroup$ @Jon I agree that common duplicates are an issue and we could indeed consider whether the stackexchange approach (choose or make a canonical version of the question and point everything else to that) is sufficient for the most popular ones. $\endgroup$ – Glen_b Jul 17 '17 at 21:59
  • 2
    $\begingroup$ I'm not sure what the best way to do something else might be. One nice thing about the current SE method is it fits with a piecemeal approach -- one person can work on making one question (and an answer) more canonical without committing to doing anything beyond that, and without having to be organized into it. $\endgroup$ – Glen_b Jul 17 '17 at 22:08
  • 1
    $\begingroup$ The Documention beta's finished on SO: meta.stackoverflow.com/questions/354217/… $\endgroup$ – Scortchi - Reinstate Monica Aug 15 '17 at 10:58
  • 3
    $\begingroup$ I'm marking this as declined because the beta has been canceled. I would like to revisit the idea if the chance arises. But it will be several years at least. Thank you for your interest, however. $\endgroup$ – Jon Ericson Sep 26 '17 at 23:59

You must log in to answer this question.

Browse other questions tagged .