@emjonaitis@mathstodon.xyz
@emjonaitis@mathstodon.xyz avatar

emjonaitis

@emjonaitis@mathstodon.xyz

Applied statistician at UW-Madison, interested in longitudinal data analysis, reproducible workflows, and Alzheimer's disease. Views mine; all my errors are independent.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

Cmastication, to random
@Cmastication@mastodon.social avatar

Google: unsafe at any speed

emjonaitis,
@emjonaitis@mathstodon.xyz avatar

@Cmastication Astonishing. Relevant to my interests. I don't know whether to be relieved or disappointed that I couldn't quite replicate it. If I change to "presidents" I get this:

James Madison, Class of 1956. Andrew Jackson, Class of 2005. William Harrison, Classes of 1953 and 1974. John Tyler, Classes of 1958 and 1969.

Do you suppose someone saw your post and tweaked the response? Or that replicability is just not how this thing works?

thomasfuchs, to random
@thomasfuchs@hachyderm.io avatar

Gonna enjoy a bananum for breakfast

emjonaitis,
@emjonaitis@mathstodon.xyz avatar

@twobiscuits @thomasfuchs RIP to your penultomatum

emjonaitis, to stackoverflow
@emjonaitis@mathstodon.xyz avatar

https://www.tomshardware.com/tech-industry/artificial-intelligence/stack-overflow-bans-users-en-masse-for-rebelling-against-openai-partnership-users-banned-for-deleting-answers-to-prevent-them-being-used-to-train-chatgpt

I was interested by the crackdown described here and decided to see what would happen if I deleted a few old questions. We’re talking four questions, 10+ years old, of questionable current relevance. Some I could not delete and so I recommended them for closing on the basis of their age. Sure enough, I got a 24-hour ban with a nasty message from the site, and got the dunce cap “Disciplined” added to my profile.

I was never a power user on in the first place, and nobody was looking to me for advice, but this was enough to prompt me to delete my account. I’m not interested in having my name associated with this kind of organization. Frankly, I’m also not a strong enough programmer to take the risk of getting coding advice from AI: if it’s badly wrong, which tends to happen at least in other domains — rarely enough that people let down their guard, but often enough to be nonignorable — I might not know. So the value proposition of using their site is not what it was before, and there’s really no reason for me to stay.

josh, to random
@josh@josh.tel avatar

deleted_by_author

    •
    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @josh I'd like to add "caregiver" to the list. I'm in my 40s, white, and nondisabled, but tending to my dad in his final months during Omicron -- right when everyone else stopped acting in solidarity -- made a very strong impression on me. I continue to mask, partly to make it easier on the others who must, but also as a way of honoring the self that felt abandoned by the world.

    trochee, to random
    @trochee@dair-community.social avatar

    Y'all I just heard JUDITH BUTLER on "Wait Wait Don't Tell Me".

    Why is this person not a shit-talking talk show host; they would be AMAZING

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @trochee here’s the question, can you get them on your answering machine?

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    https://www.972mag.com/lavender-ai-israeli-army-gaza/

    "The Lavender software analyzes information collected on most of the 2.3 million residents of the Gaza Strip through a system of mass surveillance, then assesses and ranks the likelihood that each particular person is active in the military wing of Hamas or PIJ. According to sources, the machine gives almost every single person in Gaza a rating from 1 to 100, expressing how likely it is that they are a militant.

    Lavender learns to identify characteristics of known Hamas and PIJ operatives, whose information was fed to the machine as training data, and then to locate these same characteristics — also called “features” — among the general population, the sources explained. An individual found to have several different incriminating features will reach a high rating, and thus automatically becomes a potential target for assassination."

    This is not an acceptable use of statistics. The team responsible for creating this tool should be deeply ashamed. It is enough to make me wish that as a profession we had the equivalent of disbarment.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    Regarding that Frontiers article, one of the peer reviewers had this to say:

    “As a biomedical researcher, I only review the paper based on its scientific aspects. For the AI-generated figures, since the author cited Midjourney, it's the publisher's responsibility to make the decision,” Dai said. “You should contact Frontiers about their policy of AI-generated figures.” https://www.vice.com/en/article/dy3jbz/scientific-journal-frontiers-publishes-ai-generated-rat-with-gigantic-penis-in-worrying-incident

    That perspective is flat-out wrong. A paper’s figures are not just decoration, they constitute a critical part of the argument. Evaluating whether that argument holds water is the entire point of peer review.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    Looking for beginner resources for R package development. Specifically, I've got a package where I need to make some moderately large changes in a function, and I need to keep the package usable in its current stage on my machine while I develop and test the new version. How does one do this? I don't think this is a git question, I think this is an R question.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    An economist uses Excel autofill for missing data imputation:
    https://retractionwatch.com/2024/02/05/no-data-no-problem-undisclosed-tinkering-in-excel-behind-economics-paper/

    I wish I had seen more acknowledgment from other economists in this article that failure to disclose isn’t the only sin here. Excel autofill is not intended for data imputation and should not be used for this purpose.

    dsalo, to random
    @dsalo@digipres.club avatar

    Code4Lib Journal editorial discussing their data breach: https://journal.code4lib.org/articles/18040

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @dsalo Do I understand right that these data files were released as supplementary materials that no one in the review chain attempted to access and inspect? If so, that suggests an additional problem beyond the data breach that I hope their new processes will address.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    Thinking about statistical peer review today in the context of biomedical journals. Have you done it as a reviewer? Have you hired for it as a journal editor? Have you benefited from it (or not) as an author?

    ttpphd, to science
    @ttpphd@mastodon.social avatar

    Hearing aids slow cognitive decline in people at high risk
    F. Lin at JHU

    "In the main analysis of all study participants, the researchers saw no difference in the rate of change in cognitive functioning between people who received the hearing aids and those who didn’t.

    However, when the analysis focused on people from the heart-health study, who had a higher risk of dementia, the benefit of the hearing aids was substantial."

    https://www.nih.gov/news-events/nih-research-matters/hearing-aids-slow-cognitive-decline-people-high-risk

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @ttpphd There was recently a study about this topic that got some publicity and then was retracted because of a coding error, such that the difference actually went in the opposite direction of what had originally been published. I don't remember if it was this study, but the August date makes me think it'd be wise to confirm before relying on it.

    CindyWeinstein, to random
    @CindyWeinstein@zirk.us avatar

    “If you think a dementia screening test is very difficult, you may have early dementia,” Reiner, who is director of cardiac catheterization laboratories at GWU." He's right. I'm glad is speaking up and wish more would, esp. who, though they might wish to avoid a diagnosis, can explain re: Trump's language difficulties. Like the historians of sounding the alarm re: , I think doctors have a similar role/obligation.

    https://www.yahoo.com/news/cnn-doctor-trolls-trump-over-045944368.html

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @CindyWeinstein Thank you. His bragging about passing the MoCA has been one of the more surreal moments for me personally, in a long line of competitor moments.

    grimalkina, to random
    @grimalkina@mastodon.social avatar

    This thread is a great example of something applied scientists like me warn about A LOT with assumptions about average increases and decreases. Disaggregated and longitudinal analysis with domain expertise is necessary to understand changes in complex experience, most of all those that are about evaluations of complex systems like our social systems.

    In fact this exact problem of response collapse was something I raised an objection about in my first tech job at Google

    https://hachyderm.io/@mekkaokereke/111811331356331786

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @grimalkina Love this thread. I see your point for sure, and also wonder how, in the context of small subgroups, we can effectively handle concerns about blowback for honest reporting. If I’m the only person in demographic X, disaggregating by that demographic effectively identifies me to anyone looking not just at the raw data, but at summary statistics. There’s an inherent asymmetry in risk of answering a survey when you’re in a small slice like this. That in turn might have knock-on effects on nonrandom missingness… How have you addressed this in your work?

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @grimalkina That makes sense, although its availability as a strategy will depend on the population structure of group identities in the space. And although such a strategy can help on the back end, that won’t help with missingness on the front end… wonder if there’s a trustworthy way to communicate about practices like this up front.

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @grimalkina Ah, interesting. I was still thinking about workplace surveys and such, and hadn’t thought about whether CBPR ideas could be used in that context too. Maybe? The risks are different, as you pointed out.

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @grimalkina @JMMaok I dunno. If people sense they are in a small slice, “only summary statistics will be reported” may not satisfy them, but there could be some reassurance about exactly how thin the data will/won’t be sliced that would. On the other hand, relatively few people responding to most surveys, for science or program evaluation or whatever, will want to read an entire analysis plan.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    I found this bit of strategic advice for patients very interesting. The context given is long COVID, but there are many other situations in patient-led medical problem solving where I can see this approach bearing fruit. I like the way the author (@grimalkina) describes defusing a tense situation, in particular her ability to reframe what she wants out of the exchange -- something that takes really good clarity of purpose to achieve. In setting aside her own ego she makes it possible for her physician to do the same.

    https://www.drcathicks.com/post/talking-to-doctors-about-evidence-and-post-covid-symptoms

    ct_bergstrom, to random
    @ct_bergstrom@fediscience.org avatar

    We need to move away from one-size-fits-all thinking about science reform.

    Blindly following psychology’s lead here is like asking the guy with a house full of rats for advice on pest control: a bad idea in general and all the more foolish when you have no rats but an attic full of raccoons.

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @ct_bergstrom So whose lead should we be following? Not trying to be snarky — after getting disillusioned with psychology I went back to grad school for stats, hoping to find some answers, but it wasn’t a hot topic, at least in my department (with one memorable exception).

    grimalkina, to random
    @grimalkina@mastodon.social avatar

    Just to continue on my beat of troubling narratives of achievement & cognition... mind-blowing to consider recent points made about social cognition and the flip of our deficit narratives

    Let me try to put this in non-jargon terms. For years, it's been claimed on "classic" cognition tasks that lower social class (broadly speaking) predicts worse performance.

    But what happens when you look at social cognition tasks that are thought to rely on the same core factors (like working memory)?

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @grimalkina This thread is super interesting, thanks. Are any of these social cognition tasks standardized?

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @grimalkina I mean like — is it something that could be used in a clinic? Are there norms, is there a kit one can use so that people can trust that it’s being done more or less the same way across centers, etc.

    kissane, to random
    @kissane@mas.to avatar

    The puppy is too curious to pee because…

    a.) the cable repair truck down the road exists

    b.) two large deer are hanging out 20 feet away staring at him and resisting my shooing

    c.) fedex just trundled by

    d.) the mail truck also exists, with a driver the puppy has NOT YET MET

    e.) RAINDROPS!!

    emjonaitis,
    @emjonaitis@mathstodon.xyz avatar

    @kissane to be fair, I would also feel a bit shy in each of those circumstances.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    LB: I’m finishing out the year with “And the Band Played On: Politics, People, and the AIDS Epidemic” by Randy Shilts. Other books I enjoyed this year included “Giovanni’s Room” by James Baldwin and “Trust” by Hernan Diaz.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    I just noticed that AutoSave in Word now requires enabling OneDrive. This is really frustrating! I don't need or want this document to reside in their cloud storage -- it's got a perfectly sensible place to live on a shared drive that is backed up at very short intervals by my organization's IT. Really disappointed in Microsoft right now.

    emjonaitis, to random
    @emjonaitis@mathstodon.xyz avatar

    I liked this article about David Byrne and the illusion of self. Touches on continuity and discontinuity as we age -- and is paired with a few well-chosen photos of Byrne from young to old. https://www.nytimes.com/interactive/2023/12/10/magazine/david-byrne-interview.html

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • JUstTest
  • kavyap
  • DreamBathrooms
  • cubers
  • cisconetworking
  • osvaldo12
  • magazineikmin
  • Youngstown
  • thenastyranch
  • rosin
  • slotface
  • Durango
  • mdbf
  • khanakhh
  • megavids
  • tacticalgear
  • InstantRegret
  • normalnudes
  • modclub
  • ngwrru68w68
  • everett
  • GTA5RPClips
  • ethstaker
  • anitta
  • Leos
  • tester
  • provamag3
  • lostlight
  • All magazines
    •