awesome post, and couldn't agree more with the thesis R is great because it "just works" most of the time

Job update! It has been a little over a month now, but I'm excited to share that I have joined Oscar Health, where I'm leading a team focused on our clinical risk adjustment process 🤓

wooahh! need to give this a read 🤓

simply amazing

It has been quite a while since I posted my blog on the modeling the Dunning-Kruger effect. I intended for this to be the first of a series on modeling some classic effects in psychology/cognitive science. Any recommendations for the second post in the series? haines-lab.com/post/2021-01...

last of us S2 E2, more like last episode amirite

the moment some unintuitive simulation and empirical results finally click and you know what needs to be done 🤌 a feeling vibe coders can only dream of

SIX YEARS after the initial blog post, this paper is finally published.. what a wild ride - the blog: bit.ly/3GbOqQa - original tweet thread: x.com/Nate__Haines... - published (open access) paper: doi.org/10.1037/met0...

sounds like a black mirror episode

Using this as an excuse to share a great post by @natehaines.bsky.social on modelling and Dunning-Kruger. haines-lab.com/post/2021-01...

yes Yes YES! Feel your worries melt away as you watch the "Raccoon Whisperer" feed 25 loving raccoons in freezing weather. Clip below. Link to full 20 min video: www.youtube.com/watch?v=Ofp2...

1/6 Hello Bluesky! 👋 Excited to join this community and share my new blog. First post: Using Bayesian hierarchical models to rescue "unreliable" cognitive tasks, with the dot-probe task as my case study. cogpsychreserve.netlify.app/posts/dotpro...

this album slaps

I am surprised that many data scientists (perhaps without a stats background) do not recognize that CUPED is regression. In this post, I show you why you can't spell CUPED without FWL dpananos.github.io/posts/2025-0...

My team at Ledger just released `bermuda`, an open-source library for insurance triangle data manipulation! Check out the repo and docs below to see if it would be useful for your own work: GitHub: github.com/LedgerInvest... Docs: ledger-investing-bermuda-ledger.readthedocs-hosted.com/en/latest/

Please send me your fave blog post you ever wrote. Need some inspiration

The answer to this is that you get something that looks like: results = { 10: 1.3, 25: 1.3, 50: 1.3, 75: 1.3, 90: 1.3, } Why? It is due to late binding of the iterator value (p) in the lambda function within a list/dict comprehension ☠️

must be no pythonistas in bsky huh

time for a valentines day pop-quiz! without cheating, what will the output of `results` look like?

humans put lightning inside rocks to make them count numbers and now i have to receive emails

1/3 IMO, this method gets back to a more common way of testing parameter recovery. Back in the day, I would typically fit my model to data, generate posterior predictions using the fitted posterior means, then re-fit the model to the posterior predictions to see if I can recover those means well.

In a funny turn of events, I have come to disagree with one of Paul Meehl's greatest pieces. After spending a couple years in the insurance industry, I can confidently say that actuaries use just as much subjective judgement as clinicians 🤖

lol on old twitter this post would have elicited a bunch of responses ranging from how incorrect i am, to people agreeing, to some sharing relevant resources. here i get literally nothing, this site is boring

This project made me curious—why is there relatively little research out there on quantitative item reduction strategies? There's lots of guidelines as well as heuristics, but you would think that for something so important, there would be a richer literature on best practices 🤔

Been trying to find satisfying approaches to address item redundancy and like a gift from the gods here we are

I am excited to read this

1/N Some New Years reading to share! In this post, we dive into Cronbach's alpha, Fisher info, KL divergence, and Bayes factors as measures of item informativeness. We then use these metrics to reduce a large 100 item pool down to just 15 items while maximizing information 🤖

Test-retest paper out! This one has been some time in the works (thanks to other co-authors for bearing with me as I moved institutions and then went on maternity leave). With @olijrobinson.bsky.social, Katrina Tan, Hoda Tromblee and Michelle Wing. doi.org/10.5334/cpsy...

🙄

CounterCounterpoint: Spoilers though

My PyData NYC 2024 talk is now online, go check it out! 🤖 I cover all of: - the "why" of model blending - the power of hierarchical stacking - a walkthrough of our open-source library (BayesBlend) API - a practical example using real-world insurance data youtu.be/hSGrDXJ2L7M?...

p^gamma ———————————— p^gamma + (1-p)^gamma doi.org/10.3758/s134...

I'm reviewing too many papers these days..

Computational reproducibility is a solved(ish) problem The reasons those solutions haven't permeated academic social science are legion, and a prominent one is lack of accessible training materials This post is a great intro to using these tools from someone with a social science background

While we are on the subject of fraudulent data and broken data processing + analysis lineage... check out my blog out how to use renv, Docker, and GitHub Actions to automate building computationally reproducible R projects! haines-lab.com/post/2022-01...

BayesBlend just had it's first external contribution! 🥲 I <3 open source github.com/LedgerInvest...

PyData NYC 2024 was a success! One of the most enjoyable conferences I've attended—learned a ton, and got to meet and hang out with some cool people. My talk was recorded, so look out for a post on that in coming weeks! 🤓 nyc2024.pydata.org/cfp/talk/WAW...

Heading to New York for PyData NYC! My talk is this Friday at 11:40am For any mutuals who will also be attending, feel free to reach out! Would be great to meet up 🤓 nyc2024.pydata.org/cfp/talk/WAW...

Go watch my colleague Conor Goold talk about some cool hidden markov models we have been developing at Ledger!

StanCon 2024 Oxford: recorded talks are now released! statmodeling.stat.columbia.edu/2024/10/30/s...

tragedy in three parts

JOIN US 🤖

If you think there is no tension here, just start following meta/open-science folks, might change your perspective 😁

Just wanted to note how cool it was to be able to work on this paper. People are always very surprised when I tell them how much of the Bayesian toolbox we take advantage of at Ledger, in part because there's a (misguided) idea out there that Bayes is not used in industry 🤓

I'm excited to share our Bayesian workflow on securitizing casualty insurance risk! We leverage state-space models, simulation-based calibration, prior- and posterior-predictive checks, leave-future-out cross-validation, model stacking, and more 🤓 arxiv.org/abs/2407.14666

I honestly didn't think the authors could put themselves in a worse position than they already had, and yet...

"The results look pretty replicable. I don’t think the preregistration criticism changes my mind a lot about the paper.” Do people even know how to read? It's maddening that we wrote an MA documenting problems, there are two independent blog posts reporting on further issues & it's prereg all over!

The most recent brms CRAN version added support for loo_epred() and moment matching for LOO-CV predictions, which makes it easy to make, for example, predictive probability calibration plots using the LOO-CV predictions #Bayesian #rstats

I've long wanted to write a blog post about how you can create a really solid stack that achieves most Open Science goals in a way that exceeds what you'd get from depositing somewhere like OSF. The recent *events* have driven home how much room there is for improvement in how we store things.