To clarify: if an individual moves are they still practically speaking 'at-risk' given your method of collecting data on diagnosi of A or B, i.e after moving would the diagnosis of A or B appear in your data? And can one individual experience a or b multiple Times?
I continue to be fascinated by the phenomenon whereby an expert engages with any of the LLMs on their field of expertise and is instantly horrified by the wrong answers, and then goes on to use it for things they are not experts in as though it won’t be just as bad for those.
#rstats
Help með understand the US accreditation system, when I read that someone is a "triple board certified" doctor, am I to understand that they spent roughly a decade as a resident? Or does the system allow for some kind of transfer of credits so that the next residency takes shorter?
one of the worst things you could do is solve all my problems and take from me my one true joy in life: complaining
We are about to get very cozy with in-person oral exams. Of course that will destroy university finances...
There's promise in using LLMs for code review, but it's tricky things to make sure it's not overwhelming.
I was looking at this new experimental package by Simon Couch and I really love how it allows you to review code iteratively. #rstats #ai #llms
github.com/simonpcouch/...
screenshot of the reviewer package, taken from Simon Couch's github repository https://github.com/simonpcouch/reviewer
The reviewer #Rstats package from @simonpcouch.com has an excellent idea: instead of letting an LLM rewrite your file wholesale, it proposes small structured edits with reasons, and you walk through them in a Google Docs-style track-changes UI
'How to interpret hazard ratios', with @dominicmagirr.bsky.social and @timpmorris.bsky.social thestatsgeek.com/2026/01/15/h...
v0.3.0 is now available on Firefox, soon to be available on Chrome and Microsoft Edge web stores,
Here's what's new:
You may now turn highlights into annotations!
Bundled pdf.js, to annotate PDFs, click the margin icon on your toolbar, as if it was to open the popup!
🧵
A reminder of one of the loveliest pairs of siblings in the dictionary: ‘muscle’ and ‘mouse’. To the Roman imagination, the flexed biceps of a (typically naked) athlete resembled a rodent scuttling under the skin. ‘Musculus’, in Latin, means ‘little mouse’.
are you using a DEFCON-based strategy to manage bowel motility in hospitalized patients?
having a clear, evidence-based treatment strategy improves efficacy & saves time
you don't need to redesign the wheel for every patient
& helpful to avert thermonuclear war
emcrit.org/ibcc/constip... #EMIMCC
I always find this image a bit misleading because it focus on the year studies are *published*, not when they are *started*.
Here is another version of that figure using the start year of study rather than publication year. Sample sizes in the early 1990s were larger than previous years.
You can specify relative position in ggplot with I(x) I(y).
E.g. annotate("text", x = I(.5), y = I(.5), label = "hello!") will place the text in the middle of the plot.
This, combined with alignment arguments is like 87% of the magic for me.
That is the most insane thing I've read. This is not how it works outside USA
I launched version 3.0 of my browser extension "Lazy Scholar", a free in-browser research assistant. It opens automatically when you load an academic article.
See: lazyscholar.org/2026/01/10/l...
Steph Curry Finally Finishes Eating Mouthguard
Steph Curry Finally Finishes Eating Mouthguard
I just discovered this Science & Futurism podcast by this guy named Isaac Arthur. It's just this super smart, seemingly very nice guy with a relaxing voice riffing about cosmology and physics and sci-fi concepts. He puts out an episode like every 3 days and I've enjoyed every episode so far.
Did you know? 2026 is one of the integers which, when divided by the sum of the squares of its digits, does nothing particularly remarkable
We're doing the final sprint, and I think we're able to send the PDF of the forthcoming Bayesian Workflow book to the publisher in the next two weeks (500+ pages), which would mean it would be published some time next year
Doing a bottom-to-top close edit now for errors and clarity. Just about 100 pages to go, not quite yet insane. Like some wetware LLM, I have been trained deeply on the distinctive text stylings of my individual coauthors. Feel I could produce novel utterances in any of their voices. Almost there!
Here’s the original coding line: replace event`i’ = 1 if delta_mct`i’ != 0 | spouse_delta_mct`i’ != 0 And here’s the corrected coding: replace event`i’ = 1 if (delta_mct`i’ != 0 | spouse_delta_mct`i’ != 0) & delta_mct`i’ != . & spouse_delta_mct`i’ != .
I am just learning of this 2015 retraction, adding to my "science as amateur software engineering" files. Seems they classified missing values as obs outcome of interest (divorce). Classified 32% of sample divorced, rather than true 5%. retractionwatch.com/2015/07/21/t...
Models as Prediction Machines: How to Convert Confusing Coefficients into Clear Quantities Abstract Psychological researchers usually make sense of regression models by interpreting coefficient estimates directly. This works well enough for simple linear models, but is more challenging for more complex models with, for example, categorical variables, interactions, non-linearities, and hierarchical structures. Here, we introduce an alternative approach to making sense of statistical models. The central idea is to abstract away from the mechanics of estimation, and to treat models as “counterfactual prediction machines,” which are subsequently queried to estimate quantities and conduct tests that matter substantively. This workflow is model-agnostic; it can be applied in a consistent fashion to draw causal or descriptive inference from a wide range of models. We illustrate how to implement this workflow with the marginaleffects package, which supports over 100 different classes of models in R and Python, and present two worked examples. These examples show how the workflow can be applied across designs (e.g., observational study, randomized experiment) to answer different research questions (e.g., associations, causal effects, effect heterogeneity) while facing various challenges (e.g., controlling for confounders in a flexible manner, modelling ordinal outcomes, and interpreting non-linear models).
Figure illustrating model predictions. On the X-axis the predictor, annual gross income in Euro. On the Y-axis the outcome, predicted life satisfaction. A solid line marks the curve of predictions on which individual data points are marked as model-implied outcomes at incomes of interest. Comparing two such predictions gives us a comparison. We can also fit a tangent to the line of predictions, which illustrates the slope at any given point of the curve.
A figure illustrating various ways to include age as a predictor in a model. On the x-axis age (predictor), on the y-axis the outcome (model-implied importance of friends, including confidence intervals). Illustrated are 1. age as a categorical predictor, resultings in the predictions bouncing around a lot with wide confidence intervals 2. age as a linear predictor, which forces a straight line through the data points that has a very tight confidence band and 3. age splines, which lies somewhere in between as it smoothly follows the data but has more uncertainty than the straight line.
Ever stared at a table of regression coefficients & wondered what you're doing with your life?
Very excited to share this gentle introduction to another way of making sense of statistical models (w @vincentab.bsky.social)
Preprint: doi.org/10.31234/osf...
Website: j-rohrer.github.io/marginal-psy...
the world has changed
Our guidance regarding performance measures for medical AI models is finally out!
- Stop bashing AUROC, although it does not settle things
- Calibration and clinical utility are key
- Show risk distributions
- Classification statistics (e.g. F1) are improper
www.thelancet.com/journals/lan...
My love for em dashes is as old as the hills and as mighty as my inability to make a transition between thoughts literally any other way and I will be damned if I let the clankers take that from me
screenshot of a google search for “250 ml to cups” where the results show 250 cubic miles is equal to 4.404 x 10^15 cups
thank you yes that answers my question
If you have a manuscript ready to submit in mid-December do you:
a) Submit in December (when editors are bombarded with manuscripts)
b) Hold off until mid-January to let the editors have a damn holiday already
{DAGassist} let's you process a DAG and estimate models giving {dagitty} input and baseline model #CausalSky #rstats
cran.r-project.org/web/packages... I think {ggdag} has more to offer for analysis of a DAG, while {DAGassist} is a one-stop package for classifying variables and getting estimates
