1
$\begingroup$

I'm not entirely sure how good this idea is, but basically there have been a few times here where, as a by-product of answering/asking a question, I learn about something fairly new in statistics that I wish more people would know about. Something I learned about this year was this method of making model predictions with SEM and another method I learned about a year or two when it first came out was the partR2 package for mixed models, which again came about randomly from asking questions here and on SO.

The idea I was considering is having some kind of mega-thread or wiki that updates new advances in statistics that are notable in some organized place so people who casually browse can keep themselves up to date with what is considered cutting-edge. Best would be having the statistics explained in a simple way rather than just linking whatever, but any information is better than none. Some of these things get lost under the mat and it would be good to have a resource to point people to if they want to learn what's out there. I also realize that some of these things can simply be looked up in places like Psychometrika, Journal of Statistical Software, etc. but the point is to make these techniques more accessible to a wider audience.

Feel free to shoot this idea down if it has been suggested before or is infeasible. Just a thought. I found a somewhat similar suggestion here.

$\endgroup$
10
  • 3
    $\begingroup$ I think it is more akin to a blog rather than a mega big list, imo. Interestingly, people at SciFi.SE have a blog. $\endgroup$ Apr 2, 2023 at 15:04
  • $\begingroup$ Yeah my thinking was based off what I saw on the Chinese SE before where they had some long thread that included many resources for learning Chinese. I figure something similar could be produced for yearly updates regarding interesting/useful techniques. $\endgroup$ Apr 2, 2023 at 15:14
  • $\begingroup$ Apart from what Tim articulated (+1), the main con is the dearth of readability. Think about it: every now and then a result is extended, an algorithm is modified, a new class of theorem is proposed and so on. If we were supposed to report them coherently in good spirit also, the post cannot be organized in a state subsequently defeating the original intent. $\endgroup$ Apr 2, 2023 at 15:23
  • $\begingroup$ I suppose that makes sense. $\endgroup$ Apr 3, 2023 at 12:10
  • $\begingroup$ At sister site MathOverflow there is a thread News of potential interest to the MO community, which seems to be mostly used for obituaries ... $\endgroup$ Apr 4, 2023 at 22:59
  • $\begingroup$ Jesus haha well thats something I suppose. $\endgroup$ Apr 5, 2023 at 0:47
  • 1
    $\begingroup$ A big list might not work so well. The format of q&a can become cluttered easily. Possibly an indirect pointer to other resources that discuss new developments might work easier. $\endgroup$ Apr 7, 2023 at 9:04
  • 1
    $\begingroup$ Possibly a tag that combines questions about new developments could bring new developments more easily to attention. You can search that tag to get questions about this topic. A problem might be to manage that tag. Questions about new developments might not become tagged with it, and the other way around questions that are not about new developments might become tagged with it. It is a bit subjective. $\endgroup$ Apr 7, 2023 at 9:09
  • 1
    $\begingroup$ We used to have a blog. I don't know if it could be revived. You can also post such things in chat. That does happen from time to time ('neat paper on X just came out in Journal'). $\endgroup$ Apr 8, 2023 at 12:04
  • $\begingroup$ I wonder if a tag like "new-methods-developments" might be a more viable alternative. Then a separate Q&A for each new such development, linked by the tag. $\endgroup$
    – mkt
    Apr 9, 2023 at 9:02

2 Answers 2

7
$\begingroup$

The only way we could have it is to create a community wiki question that anyone could answer. There is no blog functionality in StackExchange.com, so it could only be a question with answers. There would be multiple problems with it:

  • Anyone could answer it, posting anything since the open-ended nature of the question. We would not be able to control for duplicate entries or the quality of the answers. It would basically be a mess.
  • There is no tags functionality for answers, it would be hard to search. It would possibly have hundreds of answers that would be hard to navigate. It wouldn't be a great format for it.
  • CrossValidated.com is organized about a broad variety of topics, so likely such a thread would also be like this. It would mix niche topics in Item Response Theory with recent developments in epidemiology, ChatGPT, etc. Unless you are interested in reading completely random news on many topics at once, it would be hardly useful for anyone.

There are already many blogs, journals, podcasts, site aggregators (e.g. Reddit) etc that are much better for this purpose. They have predictable quality, are focused on specific topics, and have better UX for the purpose. I don't think that having it would be a good idea.

$\endgroup$
3
  • $\begingroup$ I think those are generally reasonable points. Thanks for providing your insight. $\endgroup$ Apr 3, 2023 at 12:10
  • 1
    $\begingroup$ My reaction to this interesting but I fear impractical idea was "blog". Your impressions may differ but mine are that blogs rose and peaked quickly before fading away from that peak. Few people have time, inclination, and enough width and depth to sustain them, even when working as a small group, not one individual. Besides, many of us would find that the latest developments were (a) not relevant to what we do (b) too hard to follow (c) superseded quickly by something else. Most people who apply statistics should not even to keep up-to-date, as it gets in the way of doing your own work! $\endgroup$
    – Nick Cox
    Apr 3, 2023 at 17:30
  • $\begingroup$ That is true. I suppose there are other more fruitful ways for people to spend their time. $\endgroup$ Apr 5, 2023 at 0:49
3
$\begingroup$

I agree with @Tim's reasoning that a "everything new in stats" thread would be unmanageable and not very helpful. Part of the problem is the open-ended timeframe.

One option might be to ask a question on the main site with a limited time-frame and perhaps topic. Probably best if asked somewhat retrospectively rather than as a constantly ongoing thread in which recent answers get buried.

  • "What were the big developments in machine learning in 2022?"
  • "What were the major statistics methodology publications in 2023 Q1 (Jan-Mar)?"

I think questions like those would be considered on-topic here, though choosing the landmark developments is obviously subjective (and easier in retrospect). I don't know what quality of answers they would bring or whether asking in a different forum would do better.


EDIT: I think @gung is right that the way I phrased those question suggestions, they may be liable to be closed for being opinion-based. So it would be necessary to tighten them up by introducing more objective criteria, e.g. asking which papers in a particular field from a particular period have been most highly cited, or (more subjectively) have been awarded a prize or cited by other researchers or textbooks as being particularly influential.

$\endgroup$
9
  • $\begingroup$ I really like this idea $\endgroup$ Apr 7, 2023 at 13:39
  • 1
    $\begingroup$ Those questions would be the way to go, but they would probably be closed as opinion based. I don't see an option within the CV system to make this work. $\endgroup$ Apr 8, 2023 at 12:02
  • 1
    $\begingroup$ @gung One option (and why I suggested doing this in retrospect, which is still useful for someone who wants to "catch up" with modern-ish developments rather than wanting to be at the very cutting edge) is to ask for most highly cited papers. $\endgroup$
    – Silverfish
    Apr 8, 2023 at 20:09
  • 1
    $\begingroup$ @gung Wikipedia also has rules against opinion-based content; they don't have an article on "the best films ever" but do have List of films considered the best. Another option here would be ask "what developments in topic X in time period Y have been considered most important?" Sometimes particular developments get praised as breakthroughs, win awards, or get cited in articles like "Ten Most Important Accomplishments in Risk Analysis, 1980–2010" $\endgroup$
    – Silverfish
    Apr 8, 2023 at 21:39
  • $\begingroup$ Those are decent suggetions, @Silverfish. I think, eg, asking for the most highly cited papers on a given topic within a given time period might work, but it still strikes me as a little dicey. I can still easily imagine such a question being shot down. If the blog could be brought back, that would probably be the best venue for this sort of thing. Otherwise, posting a link to a new paper you like, w/ a promotional blurb, is likely the way to go. $\endgroup$ Apr 9, 2023 at 0:27
  • 2
    $\begingroup$ I am struggling with this idea. I don't think it is wrong in principle but if so, why don't we have examples of such questions already? One reason seems to be that this is backwards: it is about threads people want to read, but as we all know, someone has to write them. In my own field, one review journal had a class of miniature reviews supposedly focusing on "What has happened very recently in subfield X?" but that only worked well when people subverted it. All too often papers seemed like silly articles on fashion, about what is trending up this season, usually the reviewer's own work! $\endgroup$
    – Nick Cox
    Apr 9, 2023 at 10:46
  • $\begingroup$ @NickCox (+1) Yes I remain rather sceptical of what the quality of the answers might be and I do think it would benefit from a bit of time seeing how well-used the current "in thing" ends up. Nevertheless, I'm sympathetic to someone wanting to find a way to ask, in an on-topic and productive way, "it's been 20 years since I did my stats degree and I've not really been following the academic literature, has much changed that I need to know about?" $\endgroup$
    – Silverfish
    Apr 9, 2023 at 17:12
  • $\begingroup$ As an example: it would be interesting to survey recent editions of the stalwart university textbooks and see what new material has come in that was only developed in the last X years. And conversely, what older material has been chopped in recent editions, particularly whether any methods are no longer given because they're now seen to be flawed or superseded. I wonder whether that's something which could be asked in an on-topic way, but I think it would be. $\endgroup$
    – Silverfish
    Apr 9, 2023 at 17:16
  • 1
    $\begingroup$ For other reasons I am excessively (in a Jane Austen sense) interested in the evolution of introductory texts. On present trends (mine and theirs) there is some predictable risk that I soon won't even be able to lift the 1000-page or so productions without risk of injury. Now it seems everything needs to be explained many times over and authors must include hundreds of exercises. Will no one go back to say a 300 page book that students are just expected to read, period? $\endgroup$
    – Nick Cox
    Apr 10, 2023 at 11:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .