Statmodeling Retro

As many of you know, this blog auto-posts on twitter. That’s cool. But we also have 15 years of old posts with lots of interesting content and discussion! So I had this idea of setting up another twitter feed, Statmodeling Retro, that would start with our very first post in 2004 and then go forward, posting one entry every 8 hours until it eventually catches up to the present. So far, this blog has exactly 9000 posts, so it would take a little over 8 years to catch up at this rate. But then if we continue at the current rate we’ll have another 6000 posts or so, which will take another 5 years to appear in the retro feed. Etc. So it will take awhile.

Maybe people don’t want to wait that long? We could program Statmodeling Retro to post every 6 hours, but then I’m worried that the frequency would be too high for people to follow.

Whaddya think?

33 thoughts on “Statmodeling Retro

    • “Maybe only post the most popular 20% or something so you can get thru them faster without having to post so often.”

      I think something like this idea could be a good one. Also see comment below concerning no. of comments as a criterium. Some sort of filtering like this also makes sense with regard to the worry about possibly too many posts.

      I would like to add that you could perhaps think about not making a new twitter account (seems possibly confusing, and a hassle), and simply using your normal twitter account but substituting a “normal” post with a “classic” one.

      Perhaps you could use an internet phenomenon/term for this and post the “classic” post every Thursday and make it a “Throwback Thursday” (https://en.wikipedia.org/wiki/Throwback_Thursday). You could post it on your normal twitter account by explicitly using the words “Throwback Thursday”, followed by a colon, and then the title of the “classic” post, and perhaps the date it originally appeared. Everything would kind of be self-explanatory that way as well, which i personally like (i like simplicity).

      That way you 1) do not have to set up a new twitter account, and possibly have to explain all kinds of things and/or possibly confuse folks, 2) would have a “rest” concerning having to come up with a new post for at least 1 day in the week, and 3) give re-newed attention to, and promote discussion concerning, a “classic” post.

  1. I like the idea (a lot) but what’s the reasoning behind catching up to the present? Recently Deborah Mayo posted Fisher’s 1955 “Statistical Methods and Scientific Induction” and it was for me both revelatory and a reminder that modern problems may be old ones and that recent developments sometimes make old perspectives shiny and new.

    • Very interesting point Thanatos. I would love to see the older posts. As it is, I find myself reading several older articles, the ideas of which neglected or lost. I take my cue from a point that John Ioannidis made. The fate of the Library of Alexandria is a good analogy to draw as it pertains to the disappearance of so much data and analyses.

    • …what’s the reasoning behind catching up to the present?

      I think Andrew is planing this as a huge time and labor saving device. After it catches up to the present, and goes through his backlog, he will not have to write new post for this blog any more. He will simply get them from the twitter feed. If he manages to automate that, he will be able to just sit back and enjoy his own posts without having to write or post them. Pure genius!

      • Newbie, like me, benefits from such an endeavor. I would jazz up the blog, with podcasts, interactive roundtables, music, etc. I have been learning about new communication software, which is an intriguing journey for me. I envy the ages 20-30 techies I come across b/c they are such experimenters. Today’s social media/marketing strategies are so sophisticated and save you tons of drudge work too.

  2. How about clustering topics and then releasing each topic at a much less frequent cadence (once a week). You can even provide context on how a topic evolves and discussion evolves over time. I’d look forward to the ‘well… I’d run a hierarchal model’ topic :)

    • sam’s idea is a good one, but I assume Andrew is thinking about something that doesn’t involve much new work — every hour spent on old archives is one less hour not spent on new stuff.

      Jojo suggests only posting the top 20% or so.

      Combining these ideas: maybe find the top n or x% in terms of comments / retweets or some other easy measure within each keyword, or keyword combination. So today would count under Decision Theory & Teaching.

  3. Just be sure that you don’t post on days when women are ovulating, or when there’s a shark attack, or don’t put any smile faces, nor use any high degree polynomials when fitting a model of what or when to posts :)
    Ps. I bet Bem predicted this.
    Pps. If we think of the time-reversal heuristics, olders posts may be more interesting than new ones :)

  4. Ha – this comes the day after I was trawling through the older posts to see when it was that the whole replication crisis thing started being mentioned. Then I remembered the post with the timeline of this. I like the idea and the thing is that in re-reading the old posts I got a feel for how things have changed (advent of Stan etc) yet underlying basics remain pretty much there. Not that I’ve ever used twitter – is it any good?

  5. This reminds me of an old paradox in philosophy, about the race between Achilles and the turtle. The turtle gets a head-start on the race, and some Greek philosophers concluded that Achilles, despite being the fastest runner in ancient Greece, will never catch up with the turtle, because as soon as he has reached the place where the turtle started, the turtle will have moved on to another place. And when he reaches that, the turtle will have moved on again. And so on and so on ad infinitum.

    Will your effort to have that legacy-blog Twitter thread catch up to the current state of your blog meet a similar dismal fate? ;)

    • Though a fan of Zeno of Elea, David Hilbert and Andrew Gelman (despite his tendency to integrate over the space 0-1) I nonetheless hope that statistical methodology issues will be resolved long before .9999999…. -> 1.

      • > hope that statistical methodology issues will be resolved
        Its just bad meta-statistics/physics/inquiry to hope that.

        Induction/inference/(or any scientifically profitable method for inquiry) evolves and the only assurance they ever provide is that with persistence we can expect to get less wrong (on a given question and or method of inquiry). It is only in math that things seem to resolve except for the always non-zero probability of a mistake unnoticed in the proofs.

        However, disagreements between experts can be resolved if they are given enough encouragement and incentives to discuss the differences in a scientific spirit.

        • “Induction/inference/(or any scientifically profitable method for inquiry) evolves and the only assurance they ever provide is that with persistence we can expect to get less wrong (on a given question and or method of inquiry).”

          Well put.

        • Quote from above: ““Induction/inference/(or any scientifically profitable method for inquiry) evolves and the only assurance they ever provide is that with persistence we can expect to get less wrong (on a given question and or method of inquiry).”

          Well put.”

          I never understand this reasoning. I think it might be a cop out, and/or be sort of an unfalsifiable statement, and/or be simply wrong.

          Also, how can one be “less” wrong. It’s still wrong. I can see that you could come up with a scientific example concerning some number, or quantity, that shows that someting was gradually more accurate of precisely understood or predictable, but i gather this is not applicable to every branch of science (e.g. psychology).

          To try and illustrate my point, why can’t i say the same thing (i.c. “with persistence we can expect to get less wrong”) concerning my possible astrology and tea-reading endeavours.

        • Two points:

          1. Can you give an argument that astrology and tea-reading are scientifically profitable methods of inquiry that evolve?

          To me “evolve” (in this context) means in large part that successive attempts at explanation take into account both successes and failures of previous attempts at understanding, in order to refine explanations to get a closer fit with reality. To the best of my knowledge, astrology and tea-reading do not do this.

          2. I understand “less wrong” in this context to mean “a better fit with reality” — this might involve getting numbers more precise, but it might be forming a better model that predicts more, or distinguishes between cases that the previous model failed to distinguish between, or includes possibilities that the previous model excluded. (Consider, for example medical diagnoses: A particular symptom might be involved in a variety of disorders, so diagnosis can improve by finding and including more symptoms, and looking at combinations of symptoms rather than single symptoms. )

        • +1 for the link to the Big Bang Theory quote. It puts things into perspective. (i may have a Sheldon-like brain to a certain extent).

          In scientific practice though (e.g. psychology), i still feel that there are possible problematic issues with the terms “less wrong” and (the assumption?) that there is some sort of “evolving” or “progress” over time.

          Concerning providing “an argument that that astrology and tea-reading are scientifically profitable methods of inquiry that evolve”: just to make clear, i never claimed they were “scientifically profitable methiods of inquiry that evolve”, i just want to compare them to, for instance, psychology (in light of the quote about “less rong” etc.). I am not saying you interpreted that in this way, i just want to make that clear.

          1) Now, concerning your explanation/definition of “evolve”, i question whether psychology has ever done this. Some sources would point to the conclusion it hasn’t (e.g. Fanelli (2010) “‘Positive’ results increase down the hierrarchy of the sciences”: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0010068)

          2) Concerning your explanation/definition of “less wrong”, i question whether psychology has ever done this. Some sources would point to the conclusion that it hasn’t (e.g. Hasselman (2014) “Dealing with theoretical diversity: aetiologies of developmental dyslexia as a case in point: https://osf.io/ku3c4/)

          On a more abstract level: i made the comparison between astrology and tea-reading on the one hand, and psychology on the other hand, to question the usefulness and validity of a statement like “Induction/inference/(or any scientifically profitable method for inquiry) evolves and the only assurance they ever provide is that with persistence we can expect to get less wrong (on a given question and or method of inquiry)” for studies or processes that are deemed scientific like psychology.

          Without further explanation, and examination, of the possible critical factors that would make the quote correct (e.g. stuff like your explantions/definitions of the terms, and perhaps stuff like my provided “evidence” concerning them), i reason sentences like these can be used as a cop out, can be sort of an unfalsiafialbe statement, and can be simply wrong.

        • (I replied to your comment but it hasn’t shown up yet. I put 2 links in them, and if i understood thing correctly, this could sometimes lead to a delay in publication.

          This comment here now is 1) a test to see if this one gets published immediately (like i am used to), and 2) if this is the case, to make clear my reply to your comment may take a while to get published, if at all (only if it gets rescued from the spam folder?)

        • Anon,

          I hope this is an adequate response to your 4:17 pm comment:

          “In scientific practice though (e.g. psychology), i still feel that there are possible problematic issues with the terms “less wrong” and (the assumption?) that there is some sort of “evolving” or “progress” over time.”

          To me, “less wrong” essentially means a better approximation to reality.

          I don’t mean to assert that there is *always* some sort of “evolving” or “progress” over time, but to me, *good* science aims to keep improving the approximation to reality. As for psychology in particular — my impression is that it is one of the most sloppy sciences — to the extent that I am tempted to call it a “science” rather than a science. It just seems to be less evolved (if you will) than most sciences. Of course, the study of human behavior is inherently complex (in particular, there is a lot of “between person” variability), but in my experience, psychologists all too often use very minimal evidence to. jump to strong conclusions. It’s a subject in which there needs to be a lot more acknowledgement of what is not known.

          “On a more abstract level: i made the comparison between astrology and tea-reading on the one hand, and psychology on the other hand, to question the usefulness and validity of a statement like “Induction/inference/(or any scientifically profitable method for inquiry) evolves and the only assurance they ever provide is that with persistence we can expect to get less wrong (on a given question and or method of inquiry)” for studies or processes that are deemed scientific like psychology.”

          As my comments above suggest, I think your phrase “are deemed scientific” fits psychology better than “are scientific.”

        • Anon,

          My reply to your 4:17 pm comment apparently did not get through, so here is another try:

          “In scientific practice though (e.g. psychology), i still feel that there are possible problematic issues with the terms “less wrong” and (the assumption?) that there is some sort of “evolving” or “progress” over time.”

          Good science involves some sort of evolving progress over time; unfortunately, not all science is good science.

          “1) Now, concerning your explanation/definition of “evolve”, i question whether psychology has ever done this. ”

          In my experience, psychology has seemed especially disinclined to “evolve”; it seems to be especially stuck in the “that’s the way we’ve always done it” rut. However, it just so happens that one example of likely progress appeared in my inbox today. A quote from an abstract on an FMRI study:

          “Voxel time series in fMRI are complex-valued signals consisting of magnitude and phase components, however, most studies discard the phase and only use the magnitude data. We present a Bayesian variable selection approach for detecting activation at the voxel level from complex-valued fMRI (f(c)MRI) recorded during task experiments. We show that this approach leads to fast and improved detection of activation when compared to alternative magnitude-only approaches.”

          This does sound like moving away from “that’s the way we’ve always done it” — in this case, the tradition seems to have been to discard a part of the information available from the fMRI measurements — a tradition that sure sounds like poor science.

          “2) Concerning your explanation/definition of “less wrong”, i question whether psychology has ever done this.”

          I mostly agree. There have been a few researchers who try to be less wrong, but their work so often seems to be lost in the drift of the mainstream attitudes.

        • “To try and illustrate my point, why can’t i say the same thing (i.c. “with persistence we can expect to get less wrong”) concerning my possible astrology and tea-reading endeavours.”

          If a line of inquiry isn’t profitable, you make it up on volume.

        • Qoute from above: “If a line of inquiry isn’t profitable, you make it up on volume.”

          Interesting comment in line of all the recent proposals of large-scale “collaborative” efforts in psychology.

          Also: where astrology has the daily horoscope (e.g. www (dot) astrology (dot )com), psychology may now have the “the paper in a day-approach” (http://deevybee.blogspot.com/2019/02/the-paper-in-day-approach.html)

          I think i like the daily horoscope more than the “paper in a day-approach” though. Here’s my horoscope (Scorpio) of Feb 23rd, 2019:

          “The added responsibilities you have been given lately are a sure sign that you are making a strong, positive impression on some powerful people. You are ready to take on the next challenge of life, especially if it involves travel or new cultural experiences. All you need to do right now is stick with your routine and channel any nervousness into positive energy. You simply must start believing in yourself more—after all, if you don’t, who will?”

  6. I’m with this–as long as when the stream catches up to this post, it initiates another stream that starts again from the beginning. Endless fractal of Gelman’s blogs slowly emerges from that, slowly devouring all of Internet, like Fenrir devouring the sun.

  7. I’ve not followed this blog as long as some of the others. So, I look forward to such a retro account. When I bookmark posts that interest me, I tag them along various topics and themes. Maybe the retro account could tweet out several posts grouped together along recurring themes? This might help some of us save these older posts along themes and useful tags all in one go.

Leave a Reply to Anonymous Cancel reply

Your email address will not be published. Required fields are marked *