@tslumley@fediscience.org Had my first experience with the `survey` package today. It was alarmingly easy to use, and now my little part of the paper is going to be much better than it would have otherwise. Thank you for sharing this with the world!
Remote
Erin Jonaitis
@emjonaitis@mathstodon.xyz
mastodon 4.7.2Applied statistician at UW-Madison, interested in longitudinal data analysis, reproducible workflows, and Alzheimer's disease. Views mine; all my errors are independent.
327 Followers
301 Following
33 Posts
Joined November 14, 2022
So the weird wilcox.test() behavior I saw yesterday happened because I'd unwittingly set the session to suppress all warning messages -- not just in my output, which had been my intention, but to the console. Is there a better way to handle warnings when knitting a report, such that functions that issue them frequently for ignorable things don't litter a report, but working inside a session in the console, I'll still see everything I need to see? #rstats
I'm analyzing Medicare data -- my first real experience with a large dataset, where the number of observations of interest to me is in the millions. We have repeated measures/clusters to worry about, each ranging from 2 to 10 observations, give or take.
I'm struggling with performance issues in pretty much every approach I take to this dataset. One outcome of interest is a proportion. zoib is painfully slow, even when I take a (stratified) random sample of 2% of rows -- in an hour it's only 4% done fitting my null model. Boundary values (0,1) are common in the data, ruling out "transform and just do lmer."
What general tools are available for modeling bigger datasets in R? Because of data privacy agreements I'm required to do all of the computing on-prem, so unfortunately I don't know that I can take advantage of high throughput computing on other servers, if it were even workable in this case.
More weather on the way, Madison. I hope you and yours have a safe place to be.
In a small miracle, I remembered the store of spare screens & storm windows in our basement — just in time to rummage through and replace the one that Tuesday’s hailstorm shattered. I’m not sure which previous owner is responsible, but I’m grateful.
Replying to
@grimalkina@mastodon.social wow, this is haunting and important. I'd like to boost but it's so personal I feel like I should ask first.
Replying to
@paco How sure are you that “stakeholder” means “boss” here? In my research world, “stakeholder” often means “community member” where “community” is some underrepresented group or other and the work is participatory, i.e. not fully scientist-led. Obviously you read this paper and I didn’t, of course…
This feels like a real domino falling to me. Madison rejecting Flock is reassuring but not that surprising politically, but neighboring Monona is as I understand it a former sundown town.
Replying to
@kissane@mas.to There’s a volume aspect to this too. I suspect I’m considerably more norm-core than the archetypal Mastodon user (along most axes but not all) — but because I don’t have a big account, this stuff just doesn’t come up for me.
Might be worth doing a qualitative study of pain points identified by big users of various platforms/whatever-we-call-them, and what those say about the culture. (Possibly this is already out there. Possibly you are even the person who would have written it… I’m out of my depth here) The median user of any service probably just isn’t going to experience the same things.
RE: @SeanCasten@mastodon.social
Interesting thread. This sitting Congressperson has my attention for having some idea what is actually in his purview to do.
Replying to
@tomstafford I found your essay thought provoking. The first place my brain went was yes, moving back to old signals, but the anti-meritocratic kind: sticking to the people you know, or the people who are the most similar to the people you know, with all the crassness that implies.
An extension of this to research funding which ?might? be less dysfunctional, judging by the fact that it had already been proposed anyway for other reasons -- fund people not projects. https://www.science.org/content/blog-post/funding-people-not-projects This still has potential for anti-meritocratic problems but I'm not sure that it is worse on that count than the status quo for public science funding in the US (well... status quo ante, anyway, prior to the present political situation).
I *don't* think awarding grants based on already-completed work will solve the issue, because I'm afraid research productivity is not much more difficult to fake than a grant application -- since it is also evaluated mainly via longform text. Ultimately it could be that "people not projects" founders on the same shoals.
Replying to
@CindyWeinstein@zirk.us What is potentially helpful, in my layperson view, is for leaders and others to internalize that neither his promises nor his threats can be trusted. His power comes from fear and if more people become willing to push back, that may be how we save ourselves. So I am less bearish on TACO, if that is part of how we get there.
Replying to
@millerdl@mathstodon.xyz just get a better zoom background and straighten that tie and they'll never know. Congrats!
Another github migration question, this time for the scholcomm people and librarians:
I've got a repository that is associated with a publication. I'd like to do my due diligence to keep this out of large AI models as much as I can while also keeping it available to any researchers who care to poke more deeply into what we did. (Yes, I recognize that these are in some sense incompatible goods.) Has anyone developed a standard workflow for removing a repository from github while leaving a forwarding address?
Replying to
@futurebird@sauropods.win I don't know why innumeracy in politicians makes me as mad as it does, but I'm right there with you on this
Atop this dispiriting anecdote about the state of disability services in Madison's public schools, we have quite a garden path of a headline. https://madison.com/news/local/crime-courts/article_31ab0e1d-4a8a-4ad1-b34a-847c041e9937.html
Replying to
@Paul thanks to fediverse magic, your boosted post was right above this one in my timeline, which may fit the bill.
https://mastodon.social/@dicenbuttons/116462770978339413
Replying to
@_elena@mastodon.social Je me souviens que l'algorithme de Twitter me fut sentir terrible. J'aime que Mastodon n'a pas d'algorithme; c'est moins addictif, je croix.
Replying to
@erc_bk thanks for this! I'm working on glmmTMB now and we'll see. It was tractable with a stratified sample of 25% (about 20 minutes) but 100% is slow going so far.
Replying to
@universalhub@mastodon.online @thatandromeda@ohai.social Much like insurance companies pricing in climate costs, this is the kind of bad news I love to see because it tells me the people running things have started to think this through. That said, surely science isn’t only a European affair?
Replying to
@cgbassa@mastodon.social @eldang@mapstodon.space this is breathtaking! Thank you for sharing.
Replying to
@thom @atomicpoet It was kind of a head trip returning to Twitter after a few years away, and trying to wrap my brain around what it meant that my likes might now show up in the timelines of those who followed me. That felt like a dark pattern to me, tbqh.
New to me today: The Beers List. If you have caregiving responsibilities for an elder, maybe worth a peek?
https://thecarepartnerproject.org/wp-content/uploads/The-Beers-List.pdf
Replying to
@nrennie@fosstodon.org Hah. This is my stupid human trick in real life, and the trick is exactly the same with this image. Thanks for the smile.
Replying to
@mcc@mastodon.social interesting. I’m not sure about the sensitivity of this as a marker; when I go to a repository I use a lot, I don’t see this banner, but I do see a recently edited folder called .claude — am I misunderstanding when I assume that means some of the code has been written by this tool?
RE: @luckytran@mastodon.social
I remember thinking during the height of the pandemic that the reason we needed to get comms right around Covid vaccines was not for Covid itself (as important as that was) but for the second order risk of losing MMR. But I only imagined collapse of uptake — I did not imagine we would lose it in such a literal way.
@wviechtb@fediscience.org I'm learning the metafor package for the first time this week. I'm trying to use it to compute pooled proportions across a number of small studies. Am I right that, when the input is a transformed proportion, the estimate is not back-transformed, and that I need to compute that myself?
Replying to
@erikjan I gravitated toward zoib because of its ability to handle clustered data. I don't believe that betareg has that capability. The fixed portion of the model is simple (the effects of interest are basically a group effect and a group x time interaction) but I don't want to disregard the longitudinal aspect of the data by e.g. selecting one random obs per cluster, because the trajectories are of interest to the investigators.
Replying to
@lakens@mastodon.social Interesting idea. I'm trying to think how this could work in practice. Would it solve the problem to be embedded in a scientist's lab as an RA? I think hands-on experience is valuable but I also think it would be tough to bootstrap a math person (and they're mostly math people) into designing a worthwhile study in a field that isn't theirs.
I've noticed some annoying "information leakage" between my iPhone and my MacBook Air that I can't figure out how to stop: the Safari browser on my phone interacts with the Firefox browser on my laptop even though I don't think I ever asked it to. (I figured this out through other unexpected and probably unintended behavior: using Safari on my phone automatically resizes my Firefox window.) Looking in the settings I can find, I do not see any permission for this sharing, which means I don't know how to turn it off. Any Apple users out there have advice?