And make of this what you will.
05.03.2026 11:20
👍 6
🔁 2
💬 5
📌 2
And make of this what you will.
if you're gonna write a paper claiming "LLMs can generate a statistical analysis plan for a clinical trial", and then provide THIS as an example of good output... big yikes
After nearly 400 episodes, "Stats + Stories" has become a space in which statisticians, journalists, and the curious come together to make sense of the world around them. Read this behind-the-scenes look at what it takes to produce a podcast. magazine.amstat.org/blog/2026/03... #statssky
@glenmartin.bsky.social is giving an online seminar:
"Missing Data: The Missing Component in Guidance for Developing, Validating and Implementing Prediction Models"
📅 Date: Tuesday 10th March 2026
🕐 Time: 1–2pm GMT
📍 Location: Online
🔗 Register here to receive the Teams link: lnkd.in/ejk7EFzC
I am currently running a systematic review with PROBAST-AI and not a SINGLE study has been mentioned missing data or imputation thus far.
What if you combine open datasets with AI? Apparently, a 3 fold increase in low quality research papers, mass-produced by paper mills.
Interesting study in @jclinepi.bsky.social #academicsky #episky #medsky #Skystats
Thanks to @maartenvsmeden.bsky.social for initially posting this on Linkedin!
image showing a photo of the presenter, John Jackson, ScD, and the title of his presentation, " Using Causal Decomposition Analysis to Identify Points of Intervention to Reduce Mental Health Disparities"
Join us 3/26/26 from 10:30a to Noon to hear Dr. John Jackson present, "Using Causal Decomposition Analysis to Identify Points of Intervention to Reduce Mental Health Disparities" at the next Advanced Methods for Mental Health Services Research Webinar. Register Here: jhjhm.zoom.us/webinar/regi....
The critique of unmeasured confounding is often levied in a lazy/broad way. It is trivially true in any observational study. But if the critic can't think of a plausible such confounder and posit a reasonable direction/magnitude of its bias then they're not doing productive science.
Any suggestions as to great resources (lectures on YouTube, short didactic papers etc), to teach novice researchers about RCTs? Design, endpoints, consent, eligibility criteria, IRB etc etc any and all of it.
#stats Here's a nifty interactive demo of confidence intervals by
Kristoffer Magnusson
rpsychologist.com/d3/ci/
I have been "stuck" designing studies (which mostly means back and forth meetings deliberating in our actual endpoints) for almost 2 months now...
Grading some real analysis homework and someone wrote "-M is the l.u.b. ......(of my life)" has be absolutely cackling.
Maybe its bc am over-tired from being a mentor/judge for a 48hr Med/AI hackathon over the weekend, or maybe its bc I am stuck in Ithaca for an extra 2 days due to a snow storm
There’s a lot of great reasons to use Bayesian methods (especially in industry) but none of them are the ones data influencers claim on LinkedIn 💀
My blood pressure rises bc of the "statistics influencers" on LinkedIn. Hundreds, even thousands of likes/reactions. Truly fascinating.
"I am, somehow, less interested in the weight and convolutions of Einstein’s brain than in the near certainty that people of equal talent have lived and died in cotton fields and sweatshops."
- Stephen Jay Gould
Screenshot of linked video
As statisticians, our interest in data quality is partly for selfish reasons. It's much easier to analyze carefully recorded data than the mess that some people send us! 😜
But it's not your fault if you haven't been trained. Learn more here:
statsepi.substack.com/p/simple-tip...
#rstats
The Australian and New Zealand Journal of Statistics will be running a special issue commemorating the 25th anniversary of the official release of R. (The Univ. of Auckland was the birthplace of R.) They invited various people to contribute articles, including me. 🧵 1/
Still trying to figure out their use. Extreme example: reporting age vs incidence of pregnancy for males and females combined.
I’ve never liked crude → adjusted tables. They’re not refinements of the same quantity, they’re different estimands.
Feels very related to the Table 2 fallacy (Westreich & Greenland 2013): we start treating coefficients as corrected versions of each other rather than answers to different questions.
I find it helpful to include the missing values in table 1, even if just the counts and not percentage. Granted I understand that for some nested tables it can look pretty messy and difficult to make the format elegant. Worst case: You could add a missingness attribute plot to the supplemental
A Shapiro-Wilk test of the response variable concludes very significant deviation of Normality. But residuals of linear model consistent with Normal distribution.
Visual check of the linear model with DHARMa
Periodic reminder that we should avoid testing the Normality of the response variable.
For a linear model, what matters is the Normality of residuals (and not that much). Visual checks better than test. #statistics
1/2 Nail in the coffin of dichotomania: @erik-van-zwet.bsky.social 's paper with @stephensenn.bsky.social and myself just published: onlinelibrary.wiley.com/doi/10.1002/... with extended discussion at discourse.datamethods.org/t/dichotomiz... #Statistics #StatsSky #rct #clinicaltrial
I feel like both writing and giving presentations is when I learn the most.
Wow this is a breakthrough in validation of predictive models that fully accounts for point estimates used in predictions being only point estimates. Great work @richarddriley.bsky.social and colleagues! #Statistics #StatsSky
Two survival curves. One ('late') goes all the way down, the other ('early') stays well above it all the way and bottoms out at like 30%
This is just nuts.
In a well-powered randomised trial, giving immunotherapy in the morning rather than afternoon/evening produced ludicrously big improvements in progression-free survival
www.nature.com/articles/s41...
via
www.science.org/content/blog...
"1000+ alumni publications | JAMA, Lancet, JACC" 😬
Yikes Instagram.
An Advertisement for a company that promises to learn how to do meta-analysis:
“In just a matter of weeks you can get a great study [systemic review/ meta analysis] done published in top journals…without a statistician or [slow collaboration process]” #academicsky #metascience
Look at the data! It's all over the map.
If you drink half a cup per day, your risk is a little higher than none at all.
If it's 2.5 cups, risk is 20% lower.
But if it's 4.5 cups, risk is only 13% lower.
There's no universe where this makes causal sense.
jamanetwork.com/journals/jam...
