Inclusion bias in #GWAS of #EHR traits "By weighting the sample using inverse probability weights derived from probabilities of enrollment, we replicate 54% more known GWAS variants" 😱 #statgen www.cell.com/ajhg/fulltex...

Doing statistics well is hard. Avoiding egregious errors in analysis, reporting and interpretation isn't. Just read, understand and follow the guidelines. pubmed.ncbi.nlm.nih.gov/32451178/ pubmed.ncbi.nlm.nih.gov/37286459/ pubmed.ncbi.nlm.nih.gov/40914655/ pubmed.ncbi.nlm.nih.gov/30580902/

I am grateful to Scott Berry for inviting me onto his podcast 55: A Visit with Stephen Senn: Time, Concurrent Controls, and the Bayesian Guidance www.youtube.com/watch?v=Tnb0...

Frank Harrell on why and how to do Bayes for clinical trials and the recent FDA draft guidelines statmodeling.stat.columbia.edu/2026/03/26/f...

#rstats #dataviz Titanic data examples:

I recently found out after the fact that I was the sole reviewer on a paper (major revision) and I wish I had known before the reviewer as I am not an SME on the entire paper, it makes me hesitant. Any thoughts on this? #academicsky

For anyone who is doing statistical simulation work, I honestly cannot recommend @timpmorris.bsky.social's elegant 2019 manuscript on ADEMP framework enough. Outlining a structured approach to planning sim studies, w/in-depth examples+how to display Monte Carlo SEs. #statssky#statistics #academicsky

Celebrating the draft FDA Bayesian guidance document with our perspective in @jama.com. Honored to co-author w/Jack Lee (MD Anderson), Lisa LaVange (past director of Office of Biostatistics FDA CDER&president of ASA),& my Bayesian inspiration Sir David Spiegelhalter jamanetwork.com/journals/jam...

Statistical Rethinking 2026 is done: 20 new lectures emphasizing logical and critical statistical workflow, from basics of probability theory to causal inference to reliable computation to sensitivity. It's all free, made just for you. Lecture list and links: github.com/rmcelreath/s...

I worry that a strongly informative prior, if wrong, could heavily influence results in a small dataset, whereas in a much larger dataset it would have much less effect on the posterior. I imagine you are stuck using uninformative priors? Sad to have missed the talk! #statistics #statssky #skystats

Just listened to this excellent profile of Meitner in Naomi Alderman’s Human Intelligence (the whole series is great) www.bbc.co.uk/sounds/play/...

It used to be that only at most three or maybe four people read a PhD. The author, the supervisor, the examiner(s). Nowadays even the “author” might not have read it.

"The Grad Student Who Never Said 'No'" was a blog post by Brian Wansink in which he praised a grad student for coming up with new ways to reanalyze the data. This attracted quite a bit of attention which eventually led to Wansink's downfall: www.npr.org/sections/the...

While working full-time & raising a child, Grace Wahba earned advanced degrees, completing her PhD at Stanford and becoming the first female faculty member in statistics at the University of Wisconsin-Madison. #womenshistorymonth #statwomen magazine.amstat.org/blog/2026/03... #statssky

Stats teachers, are you looking for real-world data for your students? Leticia Perez, Freida Reichsman, and Cynthia McIntyre introduce data biographies—a collection of curated data sets. Create activities for your students www.statisticsteacher.org/2026/01/29/d... #statssky

Application deadline: March 31 2026 New Researchers Conference Asia sites.google.com/uw.edu/inrc-... in Hong Kong on June 17-18, 2026, right after the IMS-APRM 2026 (IMS Asia Pacific Rim Meeting).

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

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!

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/

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 💀

"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

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.

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

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

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...

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...

You know that coffee & dementia study that's going around? It's hogwash. For starters, it's not pre-registered (that I could find), which means it may have been the product of a fishing expedition looking for dementia associations. But even if it's not ... www.nytimes.com/2026/02/09/h...

@nature.org again functioning as a marketing platform rather than upholding scientific standards. It is misleading at best to describe this as a "randomized controlled trial" AND there is no disclosure that these are Google researchers "studying" a Google product. www.nature.com/articles/s41...

the rating scale was a slider with tick marks on the 10s

Don't you f**king dare.

Quite interesting read. I wish I got to hear a little bit more about FDA decisions like this in more detail as well. evidenceinthewild.com/about/ #statssky #skystats #academicsky

It still shocks me how prevalent the Hosmer–Lemeshow gof test is, especially considering Hosmer, Lemeshow and Le Cessie published a comparison in 1997 (17 years after the HL test came out) stating its poor performance! and it has been almost 30 years since! #academicsky #skystats #statssky

"in particular mice methods tend to have the best performance. Despite their growing popularity, deep learning-based joint imputation methods, such as miwae, gain, miracle, and remasker consistently underperformed in our analysis." #statistics #statsky #academicsky #skystats

Trying to work out why all peer reviewers of systematic reviews now suggest including every sensitivity analysis ever devised (do a leave-one-out, do a sequential analysis, do p-curve, how about some outlier detection?) Is the answer ‘they are all now written by LLMs?’

This is why we need forums like bluesky and discourse.datamethods.org #academicsky #statssky #statistics #skystats

2/2 Contrast that with Bayesian designs where magic happens: If the analysis at the planned study end is valid, all interim Bayesian analyses must be equally valid regardless of the stopping rule, and they require no modification, no consideration of sampling distributions.

The estimator is fine; the assumed sampling distribution is not. If continuation depends on interim results, this isn’t a fixed-sample experiment. Reporting a conventional 95% CI implies uncond. coverage, which is false unless the futility rule was irrelevant.Would love @f2harrell.bsky.social ideas