Skip to content
 

What is the appropriate time scale for blogging—the day or the week?

I post (approximately) once a day and don’t plan to change that.

I have enough material to post more often—for example, I could intersperse existing blog posts with summaries of my published papers or of other work that I like; and, beyond this, we currently have a one-to-two-month backlog of posts—but I’m afraid that if the number of posts were doubled, the attention given to each would be roughly halved.

Looking at it the other way, I certainly don’t want to reduce my level of posting. Sure, it takes time to blog, but these are things that are important for me to say. If I were to blog less frequently, it would only be because I was pouring all these words into a different vessel, for example a book. For now, though, I think it makes sense to blog and then collect the words later as appropriate. With blogging I get comments, and many of these comments are helpful—either directly (by pointing out errors in my thinking or linking to relevant software or literature) or indirectly (by indicating misconceptions that make me realize I haven’t explained my ideas clearly enough).

But . . . I worry that the flood of topics can dilute the impact of our messages and the coherence of our discussion.

And yesterday I had an idea. As you know, I’ve recently started an On deck this week feature each Monday to get you prepared for what’s coming. It’s easy to do: I just check to see what’s on the queue and rearrange a bit if necessary.

But then recently a bunch of related items came in:

– In our recent discussion of modes of publication, Joseph Wilson wrote, “The single best reform science can make right now is to decouple publication from career advancement, thereby reducing the number of publications by an order of magnitude and then move to an entirely disjointed, informal, online free-for-all communication system for research results.”

My first thought on this was: Sure, yeah, that makes sense. But then I got to thinking: what would it really mean to decouple publication from career advancement? This is too late for me—I’m middle-aged and have no career advancement in my future—but it got me thinking more carefully about the role of publication in the research process, and this seemed worth a blog (the simplest sort of publication available to me), which I’ll post on Friday: What if I were to stop publishing in journals?

– A professor of history shared with me a story of an academic administration that seems so have purposely looked away from a case of plagiarism. I’ll discuss it this Wednesday in Plagiarism, Arizona style.

– Thomas Basbøll pointed me to a discussion on the orgtheory blog in which Jerry Davis, the editor of a journal of business management argued that it is difficult for academic researchers to communicate with the public because “the public prefers Cheetos to a healthy salad” and when serious papers are discussed on the internet, “everyone is a methodologist.” The discussion heated up when an actual methodologist, Steve Morgan, joined in to argue that the salad in question was not so healthy and that the much-derided internet commenters made some valuable points. The final twist was that one of the orgtheory bloggers deleted a comment and then closed the thread entirely when the discussion got too conflictual.

This one is a particularly rich source of material, but on Tuesday I’ll be focusing on some particular claims being made about the stringency of peer review: Literal vs. rhetorical. Then on Saturday, I’ll return to the meta-topic of the discussion with Disagreeing to disagree.

– I had a brief email exchange with Jeff Leek regarding our recent discussions of replication, criticism, and the self-correcting process of science. A key point of disagreement: what should we do about bad research that gets publicity? Should we hype it up (the “Psychological Science” strategy), slam it (which is often what I do), ignore it (Jeff’s suggestion), or do further research to contextualize it (as Dan Kahan sometimes does)? I’ll cover this on Thursday, with How much time (if any) should we spend criticizing research that’s fraudulent, crappy, or just plain pointless?

– Chris Chambers pointed me to a blog by someone called Neuroskeptic who suggested that I preregister my political science studies:

So when Andrew Gelman (let’s say) is going to start using a new approach, he goes on Twitter, or on his blog, and posts a bare-bones summary of what he’s going to do. Then he does it. If he finds something interesting, he writes it up as a paper, citing that tweet or post as his preregistration. . . .

I think this approach has some benefits but doesn’t really address the issues of preregistration that concern me—but I’d like to spend an entire blog post explaining why. This one I’m planning to post next Monday (that is, a week from now) under the title, Preregistration: what’s in it for you?

So here we have it, a week’s worth of posts on related topics. I’ll whip them all up and bump the currently-scheduled material to April.

The weekly time scale

But this got me thinking about a more general issue: what is the natural time scale for a blog? Or, at least, for this blog? Here I’ve been influenced by the writings of Thomas Basbøll, who keeps banging on the idea that there is a natural length to express any particular idea. (His magic formula is that a research article should have 40 paragraphs.) The trouble with a new topic every day is that, a day later, the last subject is largely forgotten. It all blurs in the minds of the readers until there’s a vague sense that I’m writing about some mushy mix of Bayes, plagiarism, and the voting patterns of rich and poor. Paradoxically, all the day-to-day variation can make our material seem less varied and more like a porridge.

So I like the idea of a weekly theme. And I can get away with this, because I have enough backlog that I can put together thematic weeks out of available material.

My intuition is that the week is the right time scale for what I’m trying to do here. Monthly would be too long, I think—nontechnical readers would tune out after a month of posts on statistical computing, politics-haters would be bored with a month on elections and voting, and so forth. But a week, that should work. We should have enough variation within each week to still make things interesting. I’m still talking about 5 or 6 separate posts that share some common theme (I’m thinking that weekends will remain as wild cards), not one long post divided into 5 or 6 pieces.

21 Comments

  1. MK says:

    I don’t think you can post too often, if you have something real to say – as you do, Sir.

  2. Anonymous says:

    So now instead of wasting one day at work I’ll waste a whole week when the topic is right. =-O

    • Steen says:

      I agree – it seems like there might be a lot of variability in how interesting a given reader finds any post. E.g. I want more posts about serial dilution assays. Plugs for that paper seem to have dropped off.

      Andrew, why do you thinking that posting twice as often will halve the interest in each post?

      I like this extended preview format and the weekly theme idea! To me part of the theme this week seems to be this: how can we best document research/thought processes? I’ve always thought that the ‘reproducibility/replicability crisis’ is really just a reflection of poor and inconsistent documentation standards. People don’t generally want to ‘reproduce’ work, they want to comprehend it. As far as I’m concerned the “single best reform science can make right now” is for journals to insist that far more documentation and raw data be posted online (not merely available on request) as a condition for manuscript acceptance. This would have the side effect of making people keep better records in the first place. In some cases it make things easier for people looking to “slam” or discredit research, but in others the transparency would lend credibility to the work. I wonder which would be the case for the papers that Lior Patcher has been blogging about or the case of Tyrone Hayes. http://www.newyorker.com/reporting/2014/02/10/140210fa_fact_aviv?currentPage=all

    • Steen says:

      I agree – it seems like there might be a lot of variability in how interesting a given reader finds any post. E.g. I want more posts about serial dilution assays. Plugs for that paper seem to have dropped off.

      Andrew, why do you thinking that posting twice as often will halve the interest in each post?

      I like this extended preview/weekly theme format.

      • Andrew says:

        Steen:

        I haven’t done any work on dilution assays recently but we have been thinking recently about the “cut” function, and this did arise in the assay problem. So maybe it will come up again.

  3. Robin Morris says:

    In advance of your Friday post (“What if I were to stop publishing in journals?”) can I point you to a post by Don Geman where he gives a list of reasons why conference papers should be abolished, and then advocates limiting everybody to 20 lifetime publications. How many important ideas can each person have? Warren Buffett argues similarly in terms of investing ideas.

    http://www.cis.jhu.edu/publications/papers_in_database/GEMAN/Ten_Reasons.pdf

    • Andrew says:

      Robin:

      Let me just say right now that I think it’s ludicrous to limit anyone to 20 lifetime publications. I’ve had a lot more than 20 good ideas! If Don Geman wants to restrict his output to 20 publications, that’s his choice. But I really don’t want him or anyone else to muzzle me.

      • Robin Morris says:

        I think his main point was that a paper should be complete, and was arguing against the “minimum publishable unit”.

        I had a postdoc advisor who published one paper every 4 years or so. And each of those papers was recognized some time later as being 3-4 years ahead of their time, and became foundational in their fields. But as a postdoc, trying to follow that pattern would have been career suicide – what worked 30+ years ago wouldn’t have worked 15 years ago, and certainly wouldn’t work now. But I see no shame in being known for 3/4/5 seminal papers, rather than 40 or 50 that have been cited only a few times each.

        • Andrew says:

          Robin:

          1. I have mixed feelings on this. In some cases a long paper is good, but in other settings it is helpful to have a short crisp paper clearly presenting a single idea. Indeed, I think some of my papers have been too long and would’ve benefited from being split up. In other cases, sure, there were ideas that weren’t really worth publishing.

          2. I’m sure that, for some people, 3 or 5 papers are fine. It might depend a bit on what field you’re working on. If you’re Andrew Wiles proving Fermat’s last theorem, sure. But for someone like me working in highly collaborative fields, it can be useful to put out a paper with a clear idea that others can then use. To put it another way, sure, I could pick my 20 favorite papers but that would leave out lots and lots of others that I think have made important contributions.

          • Rahul says:

            Andrew:

            You may have had ” a lot more than 20 good ideas” (you think!) but the question here is about the tradeoff between how much crap one must read before one finds what one needs.

            I think the signal to noise ratio of contemporary publishing, at least in my fields, is pretty poor. I remember being often frustrated having to read paper after paper, that looked promising from the abstract, but ended up containing half baked, or low quality, or methodologically flawed stuff.

            So, maybe there’s a case to be made to raise the bar? Can’t ideas, small analyses etc. be communicated by personal blogs or something similar? And keep the journals for the more polished, more important, really useful stuff?

            • Nony says:

              I am in favor of LPUs. Divides things up better and you go more in depth.

              Also, publish everything you do. Why not get the $$$ spent on research in some format that others can access.

              • Rahul says:

                The DoE funded projects already demand yearly reports that are posted online for open consumption. But I don’t think every funded project merits a publication.

        • K? O'Rourke says:

          I think the important part – is to _not_ get too far ahead of your colleagues so they can’t comfortably grasp and make your ideas importantly less wrong (and I believe previews are helpful here to lessen the distance).

          That should be the _purposeful publishable unit_ (of course stolen from CS Peirce.)

  4. D.O. says:

    FWIW I prefer “porridge”. There is something nice to come to a blog and be (somewhat) surprised. My impression from the Recent Comments bar on the right that on some topics the discussion lingers for much more than a day. The drawback of thematically linked back-to-back posts is that commenters might be (and will be) confused about where the really interesting discussion is going.

  5. Anonymous says:

    Regarding that preregistration comment. I think there’s something fundamentally wrong with that logic which I’ve been trying to put my finger on lately. Preregistration often goes hand in hand with this obsession with multiple comparisons, that somehow whether we decide to “look” can somehow change the nature of reality.

    For the exact same set of 1000 comparisons, people treat 1000 researchers each looking at 1 comparison differently from 1 researcher looking at 1000 comparisons. Everyone always point out that multiple comparisons are needed in the latter case but never comment on the former case even though they’re mathematically equivalent. This seems like insanity to me.

    I think often what’s posed as a multiple comparison problem is better understood in terms of base rates. When one is less selective about the correlations being examined, you’re drawing from a sparser pool of positive findings. Thus it’s the base rate that needs to be accounted for, not the number of comparisons.

    • Andrew says:

      Anon:

      You write, “For the exact same set of 1000 comparisons, people treat 1000 researchers each looking at 1 comparison differently from 1 researcher looking at 1000 comparisons.”

      I disagree. I think it is well understood that these are two sides to the same coin. The first issue is called the file-drawer problem; the second is called the multiple comparisons problem.

      One thing people don’t always realize, though, is that the multiple comparisons issue can arise even if there’s only one study and the researcher only looks at one comparison and the research hypothesis was stated (or, as Neuroskeptic might say, tweeted) ahead of time! The point here is that a analysis can be highly contingent on the data (thus causing a multiple comparisons problem) even if only one analysis was done, because the researcher has the freedom to choose the analysis (and also various data-inclusion and data-coding rules) after seeing the data.

      • Anonymous says:

        Lip service does get paid to the file drawer problem, especially on blogs and blog commentary. However, while some fields obsess over bonferroni/BH multiple comparison adjustments, nobody has ever made a “file-drawer multiple comparison adjustment” to their published p-values. This inconsistency in the operationalization of multiple comparisons indicates to me that the application of “multiple comparison corrections” doesn’t reflect a coherent statistical approach.

        But what if we really took the logic of multiple comparisons to its logical conclusion? We could enforce science-wide file drawer multiple comparison adjustments… how many comparisons would this consist of? We’d probably just end up with a literature of underpowered studies.

        In the end, I’ve come to agree with those bayesians who think multiple comparisons adjustments are a dead end. I do think they “work” (i.e. sometimes produce reasonable results) by being a roundabout way of calibrating p-value thresholds for low base rates and lack of shrinkage. I agree with you that researcher degrees of freedom and the “garden of forking paths” is yet-another form of multiple comparisons, and I definitely agree with some of the recommendations. My point about rethinking multiple comparisons as a base rate (and shrinkage) issue is that they probably are better addressed through multilevel modeling.

  6. Meh says:

    I would keep it down to 1/day and divert more of your energy into essays, books, or journal pubs. These posts are pretty interesting and it’s hard to keep up with too many of them. Maybe prioritize the more general interest ones and do less of the geeky ones. More Freakonomics, less Stan.

    That said, this thing is really first class. Reminds me of Volokh or 538 blog. But I worry about you diverting too much time into the Internet and not being well paid for it. Internet discussion can be fun, but it can have a negative video game addiction character to it also.

  7. I like the current format of diverse themes on a week. And I don’t like the idea of moving for a unique theme along a week. The current format works for me because I use feedly, and then I can read the posts I’m feeling in the mood to read. Having to wait one week for a theme I like doesn’t sound good to me.

    Maybe I’m just used to the current format and this new format will work great. I’d just like to add here that one of your readers like the current format.

  8. […] I’ve given up on theme weeks. I have enough saved-up material to do it, and it wouldn’t be too much trouble to group the […]

Leave a Reply