|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Above my pay grade for the butterfly but good luck. Glad you've enjoyed the book so far.
0 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman@jbakcoleman.bsky.social |
1260 followers 325 following 457 posts
Associate Research Scientist at Columbia journalism. Harvard BKC affiliate. Comp. soc. science, collective behavior, stats
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Above my pay grade for the butterfly but good luck. Glad you've enjoyed the book so far.
0 replies 0 reposts 0 likes
|
Dr Mircea Zloteanu 🌻
@mzloteanu.bsky.social
[ View ] |
#statstab #133 p-Curve and p-Hacking in Observational Research
Thoughts: "p-values in observational research may need to be empirically calibrated to be interpretable", or you know, stop using them in obs research 🤷♂️
#research #pvalues #observational
journals.plos.org/plosone/arti...
1 replies 2 reposts 3 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I think this is a problem with meaning two or more different things when scientists say false positive. Ioannidis has definitely made this mistake before!
1 replies 0 reposts 2 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Honestly any ioannidis paper that isn't about his CV or COVID is a breath of fresh air.
1 replies 0 reposts 2 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Honestly someone putting out that many papers has to eventually have a good take.
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
2 replies 0 reposts 6 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
With enough within-subjects samples, that assumption of normality becomes very Principle of Maximum Entropy. Consider me unprovoked.
0 replies 0 reposts 2 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Really formative paper for me seems to have fallen. Imo this is an example of good sleuthing, measured critique, and raises important questions for reconciling visual odometry research following this paper.
arxiv.org/abs/2405.12998
0 replies 2 reposts 3 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Oh that's fascinating. I wonder what other major centers/institutes are built on Meta money.
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Important read on Meta science, metascience, and how Meta has leveraged academic collaborations for PR purposes. Against the backdrop of these industry friendly findings and a collaboration with CoS, Meta is clawing back independent researcher data access.
www.techpolicy.press/the-politics...
1 replies 3 reposts 9 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Great use case for selenium, maybe with a bit of proxy magic but I'd bet they're not that secure.
www.selenium.dev
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
HOW DID YOU KNOW
0 replies 0 reposts 2 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
As it should be! A rejection with zero suggestions for improvement forces the editor to disclose the content or gives the authors nothing that can be done to improve the piece. I've mostly used it for things like "the math is beyond me make sure you have a very good physicist on this one".
0 replies 0 reposts 2 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
The other reviewer just pulled a "you didn't cite me, reject" so this was a memorable low in peer review quality.
0 replies 0 reposts 10 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
No comments to the author but private comments to the editor to reject the paper is a first for me.
6 replies 0 reposts 13 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Oh I don't mean these kinds of questions! It's when they want to say "did not respond to requests for comment" and to establish that they just bombard you with 20-40 substantive questions and say they need answers by EOD.
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
There's also the related thing.. when a Journo emails you too many questions to answer with a *very* tight deadline... they're on some bullshit to be avoided.
1 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
This is seriously the most important advice you can get as a scientist interacting with the press. If it feels seems even a little weird, decline or (if you must) answer in writing.
1 replies 1 reposts 15 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Fun fact: about half of grocery store chains have the local area code + that in their loyalty databases.
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
This is why we note that there are plenty of reasonable alternative definitions of false positives but they're not the definition of which key claims about the replication crisis have been made.
0 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
To your point above, yes it is a sign error if they are using a directional test and I think that's more reasonable definition than a point null. At the same time, models of how often a true hypothesis is tested or whatnot assume a point null and a lot of inferences flow from that.
1 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I fully expect some to show as 100% looking consistent with a false positive because of the small sample size and how binomial p values work (i.e., if I get zero out of five the p value is 1).
2 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
This is awesome. Let me check things on monday and maybe we can move to email? Fridays graph was quick and dirty so you're probably right. I'm also going to generate the graph from our posterior retrodictive to see if these results make sense with our model.
1 replies 0 reposts 1 likes
|
Berna Devezer
@devezer.bsky.social
[ View ] |
Good news everyone! The mom and pup are alive and well, enjoying a beautiful day in the water 🥰
3 replies 1 reposts 42 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I'm not sure exactly which one it was, but my guess is that it was a failed replication due to a reversal in the direction.
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Glad you enjoyed! He has a levity to his writing that is just so hard to do earnestly.
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Yeah, Henry was truly a rare breed. The book of his he spent 30-40 years working on, and shortly before it was published he joked to us that he would die before it ever got published... In true henry style, he was right!
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
For reading? Statistical rethinking, life in moving fluids, and complexity. Also calling bullshit.
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Henry horn, the author of social butterflies has the most impact on me. He was a dear friend for four years at every Friday beer hour. This video will convey who he was better than anything.
youtu.be/W9KCcN3pJec?...
2 replies 1 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
To be clear, I'm not going anywhere anytime soon... A little stint at UN things than ideally back into the academy.
0 replies 0 reposts 2 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Moved my stuff out of my office today, entering a strange interstitial place between one position ending and hopefully the next. The stack of books on my coffee table awaiting a home on a shelf was a great reminder of how awesome the past decade in science has been.
6 replies 3 reposts 44 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
This is cool and a great example of priors being a useful feature of Bayesian Inference
python.arviz.org/en/stable/ap...
0 replies 1 reposts 1 likes
|
Carl T. Bergstrom
@carlbergstrom.com
[ View ] |
Simple question for those who want to more aggressively criminalize scientific fraud. How would you prevent such systems from being weaponized against climate scientists, public health officials, etc. who report inconvenient truths? Even the threat of such abuses would be chilling.
11 replies 22 reposts 140 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Holy shit that Pleco + Neotropic. I'm blown away by this collection and utterly flattered it even reminded you of my account.
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Tentatively it *should* look something like this. Much more consistent with ML2 Table 2... ~4-5 hypotheses are consistent with FP properties but *many* of those that failed to replicate are not.
1 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I think this changes things quantitatively but not qualitatively (i.e., some of the hypotheses are consistent with being a FP but it isn't sufficient to explain low replicability). Huge catch and very important for the paper. Can't thank you enough!!!
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Very long story short, our model assumes a one-sample z-test and I coded our estimation of z accordingly, which winds up anti-conservative relative to SE for a two-sample z-test. It better reflects our theoretical model but *not* the data.
1 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Ok so it is a bug on our end. Figure 1 is less extreme than it seems, gonna rerun everything and see what we were wrong about. Thanks for flagging! Hell yah preprints.
1 replies 0 reposts 2 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Ok so it's not the easy thing (a bug in my code). That means it's the hard thing, either L&H are wrong, ML2 is wrong, or the z tests are behaving much different from the t-tests. Definitely a less conservative test but the gaps are big. This is gonna take a bit.
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Good question. L&H converts to d, and we use that with z to standardize across everything so it could be a few things. Lemme dig and get back to you.
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Banger
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I think I have one non spam account blocked and it's not even him. I do find some comfort that he still makes an effort to read my content on his alt.
1 replies 0 reposts 2 likes
|
RM (they/them)
@economeager.bsky.social
[ View ] |
2 replies 2 reposts 19 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I mean, @betanalpha.bsky.social
1 replies 0 reposts 5 likes
|
Kate Starbird
@katestarbird.bsky.social
[ View ] |
Murthy v. Missouri decision: SCOTUS rejects arguments that government illegally "censored" protected speech, ruling that plaintiffs lack standing. Very good day for social media moderation and for researchers caught in the crosshairs of a ludicrous accusation. www.supremecourt.gov/opinions/23p...
2 replies 45 reposts 125 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I have another paper hung in limbo that touches on exactly this.
1 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
GPS and accelerometer, not acoustic!
0 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
I have a dear friend who used to put acoustic tags on blue whales. A single sample might have meant days on a boat both to place and collect the tag when it floated up. A far cry from bumping a number by one and paying $5 more to a survey provider.
2 replies 0 reposts 1 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
Simple in the sense of not not requiring better theory and ideas, improving inference etc.. just doing *more* of the same thing. We don't explicitly consider resource costs... yet...
0 replies 0 reposts 0 likes
|
|
Joe Bak-Coleman
@jbakcoleman.bsky.social
[ View ] |
My early work was on sorting out why fish swim upstream and the fish kinda do the same thing today they did 120 years ago. Tbh probably the same thing they did millions of years ago in this regard.
0 replies 0 reposts 1 likes