Robin Blythe

Sign of a great talk: 80% of people coming up to you after congratulate you on a wonderful presentation. The remaining 20% tell you to eat shit and die, and that you will set back the profession by decades.

You know you've effectively been stats-pilled as an economist when you convince yourself that stochastic utility models are the most elegant explanation of human behaviour #econsky

Give this social media employee from the Dutch National Institute for Public Health and the Environment a raise! Vaccine-skeptic, in response to information about yearly flu-vaccinations: "Up your a***" RIVM: "Vaccinations are typically injected into the upper arm."

Will be speaking on Monday at the Asia launch of the Lancet Commission on Oral Health in Singapore. Come watch me tell a bunch of dentists that most dental care is just provider induced demand! www.ndcs.com.sg/events/healt...

Work-issue laptop #rstats workflow: open file explorer. Wait 3 minutes for it to crash and restart. Open RStudio. Wait 5 minutes to load. Load rstanarm. Wait 5 minutes to load. Run brms code. Wait 10 minutes to compile. Wait 30 minutes for sampling. Home laptop workflow: everything done in 3 min 🫨

Smoking-gun evidence of socioeconomic factors affecting dental general anaesthesia rates at an industrial level, and am onto my 8th desk reject because the methods are 'too simple' (comparing two logistic models + g-computation). How do I sex this thing up? Do I call it AI? Causal?

The hardest part of working on oral health grants is that you realise most journals do not give the remotest shit about dental outcomes lol. It's tough out there

Latent class analysis is astrology for scientists #statsky

It's not just your intuition; a lot of papers about LLMs are cowboy science. Important work showing that this extends to clinical trials. #medsky

Submitted to @plosone.org in March 2025. Took until Jan 2026 to get first round of reviews, I resubmit, and it goes out for review again. Now the editor is unreachable, paper's reassigned, and I have no idea what the status is. Do I cut my losses? It's infuriating. #academia

Nice study that I don't fully agree with, but makes a lot of important points about the shit show that is clinical data and the folly that is assuming that clinical prediction models work. Many of the same data issues also in clinical target trial emulation. #medsky #episky

Implementation scientists when you ask whether their frameworks or results can be quantified, tested or in any way validated #medsky ;)

My brother's nurse at the flu shot appointment: "The most common question I get these days is whether there's AI in the vaccine"

Are there any good reasons for people to take photos of slides sharing yet-unpublished research at conferences? I've now witnessed two cases of research theft from colleagues via this behaviour and can no longer assume people who photo your slides do so in good faith. Too uncharitable?

Few things more terrifying than overhearing a med student say "who cares bro, just ask Claude" regarding the curriculum

Another nice blog post (not just because I agree with it!) that reinforces a particularly tough pill for many scientists to swallow: there are lots of really strange assumptions that do the heavy lifting when we seek to call our process "exploratory", often at the expense of theoretical grounding!

Maybe a hot take, but IMO the best thing about propensity scoring is that it makes clear how absurdly difficult it is to reduce complex patients/clinician+patient interactions down to a handful of variables. It virtually always obviates some missing and/or irreducible context. #statsky #episky

Every time I forget to load tidyverse last

Actually sort of glad my paper got desk-rejected by Nature Medicine because it means I can keep shit-talking Nature

How should national expanded carrier screening programs for rare diseases be structured? Evidence from Singapore. First time running a discrete choice experiment and got to do some neat policy analyses. Shoutout to the MVP {logitr} (and, as usual, {ggplot2}) #econsky rdcu.be/6TvA8v58DKpa

Singapore: Beware of scams! Here's an anti-scam PSA you'll get multiple times a day! Also Singapore: all private companies are allowed to use marketing tactics undifferentiable from how scammers operate. We have no plans to regulate this.

This was fun. Would have also squeezed Disco Elysium in for a top 10, and this doesn't include what is probably like >8000 hours playing Super Smash Bros and Melee at friends houses

Applied #statsky and health #econsky friends: any good methods-focused conferences you recommend? I am feeling a little deflated by how little my peers in health economics care about methods. Extra points if there is a focus on Bayes, prediction, and decision theory. I'm in Singapore but can travel!

Just got a review request from a Frontiers journal that was so intensely personalised to my prior publications (in other journals) that it borders on stalking. A somewhat creepy experience.

I would like to announce I finally watched the Odyssey and quite enjoyed it, and require no further Discourse at this time. Thank you

Can we talk about something other than AI for, like, 10 minutes

This is why it's incredible to me that the US govt is not taking more charge of how these companies interface with their allies. It's using OpenAI and Anthropic in an almost mercantilist way, at least rein in their stupidest behaviours! I guess you need actual experts in charge for that, though.

Any excuse to post my favourite LotR meme

It is SO EASY to specify mo() for an ordinal variable in brms that I don't think I can go back to treating them as continuous or dummy vars. Even splines are super easy. Next on my wishlist is that mi() integrates cleanly with mo() and s() 😍 #rstats

All fields tend to some degree to confuse normative statements for scientific rigour, but I don't think I can tolerate reading any more "Planetary health" papers

Just as the AI hype types are (poorly) reinventing statistics, the anti-AI crew are (poorly) reinventing decision theory and economics. Please just open a book.

We must pace the frontier of logistic regression. It is too powerful, and the risk to humanity of incorrect interpretations of odds ratios is an existential threat #statsky #ai

Been teaching grad students a bit lately, and hoo boy, nothing makes you question education quality as much as your complete lack of confidence in your own knowledge.

I will be looking for a PhD student in Bayesian statistics (in collaboration with Paul Bürkner), located in Prague. Forward to students if you have somebody who could be interested. www.martinmodrak.cz/grammo-call/