New presentation on things we often forget to consider when designing a clinical trial: https://www.fharrell.com/talk/qrct/ #Statistics #clinicaltrials
Omics provides infinite correlations and zero causality.
Measure a million variables and finding a “signal” is trivial.
Biology doesn’t care how big your dataset is if you can’t explain WHY.
Data is cheap.
Mechanism is earned.
Gratitude to @f2harrell
R to @f2harrell: There is also an "st" command line option to get a symbol table at the end of the R file listing. #RStats
New Python script for pretty-printing of R code files - https://github.com/harrelfe/rscripts/blob/master/prprint.md #rstats
I have a new favorite editor/IDE: Zed. Open-source, blazing fast, built-in AI operability, easy to use. Heard that lots of developers are dropping Visual Studio Code for Zed. Has R and Quarto extensions. http://zed.dev
I'm surprised this passed peer review at @AnnalsofIM : https://www.acpjournals.org/doi/10.7326/0003-4819-146-12-200706190-00009 . The paper only points out the hazards of omitting Bayes from the #Statistics curriculum. Posterior means are perfectly calibrated no matter what the stopping rule. The paper deals w/sample means.
Causal by Design: new blog article on Statistical Thinking. I'm sure @yudapearl will find something to dislike but I hope there's something he likes too https://www.fharrell.com/post/causal/ #Statistics
1-day online course May 11: Intro to R/RStudio & an efficient comprehensive R workflow, rms package, enhancing regression skills to be ready for the RMS 4-day course. Details®istration at http://hbiostat.org/course @AmstatNews @InstatsS #Statistics #rstats @VUDataScience
Regression Modeling Strategies 4-day virtual course May 14-15, 18-19. More than regression: principled flexible modeling, accuracy, interpretation, selection, stability, validation, Bayes, contrast w/ML & more. https://hbiostat.org/course @AmstatNews @InstatsS #Statistics #rstats
#Statistics thought of the day: Type I "errors" in frequentist stats have always misled. The "error" exists only inside a stipulated world, not the real one. You cannot call something an error by assuming the very fact that would make it one.
Just got this email out of the blue. Didn't know ghost workers was a thing.
Nice paper on effective sample size for Kaplan-Meier survival curve estimates: https://www.tandfonline.com/doi/full/10.1080/00031305.2025.2542390#d1e2236 #Statistics
Wonderful quote that classical statisticians who will always be critics of Bayes should read every day. @PavlosMsaouel
Celebrating the draft FDA Bayesian guidance document with our perspective in @JAMA_current. Honored to co-author with Jack Lee (MD Anderson), Lisa LaVange (past director of Office of Biostatistics FDA CDER and president of ASA @AmstatNews ),& my Bayesian inspiration @d_spiegel
Thrilled to be interviewed by imminent cardiologist and clinical trialist Mike Gibson yesterday about the new draft Bayesian guidance at FDA: https://clinicaltrialresults.org/dr-frank-harrell-and-dr-c-michael-gibson-talk-about-the-fdas-new-draft-guidance-on-using-bayesian-methods-in-clinical-trials-and-it-could-be-a-game-changer/ #Statistics #bayes #pharma #clinicaltrials #rct @CMichaelGibson
Fantastic new paper casting doubt on explainability of explainable AI: https://academic.oup.com/ehjdh/article/7/3/ztag038/8512917 @MaartenvSmeden #machinelearning #Statistics
Wow! After Donald Trump mocked Gavin Newsom and other people with dyslexia and learning disabilities, Newsom’s wife, Jennifer Siebel Newsom, just completely torched him.
The end is the best part.
Correct interpretation of p > 0.05: The current sample size did not yield sufficient evidence to reject, at the incredibly arbitrary 0.05 level, the supposition that the treatment does nothing, assuming data model M (describe M ...) which forgot to include covariates #Statistics
Oh my goodness how did @JClinEpi allow this? The authors are pretending that point estimates are true values! For this study the subjects should not have been given point estimates & you really can't answer the survey questions without Bayesian posterior probs. #Statistics
One of the greatest lectures ever.
NVIDIA CEO Jensen Huang just said the quiet part out loud about what the education system will never admit.
For a century, we built humans to think like calculators.
The algorithm made that skillset obsolete overnight.
Huang: “The definition of smart is somebody who’s intelligent, solve problems, technical. But I find that that’s a commodity. And we’re about to prove that artificial intelligence is able to handle that part easiest.”
Software engineering was supposed to be the safe play.
Superintelligence cleared it first.
The SAT was supposed to measure intelligence. It was measuring the ability to follow instructions. Raw technical processing isn’t a competitive edge anymore. It’s the floor the machine stepped over before you woke up.
The question isn’t what you can calculate.
It’s what you can see before the data shows up.
Huang: “People who are able to see around corners are truly, truly smart. And their value is incredible. To be able to preempt problems before they show up, just because you feel the vibe.”
That vibe isn’t magic.
It’s the collision of first principles, human empathy, and lived experience no model can fake.
Huang: “That vibe came from a combination of data, analysis, first principle, life experience, wisdom, sensing other people.”
The operators who see around corners will command the AI.
The ones waiting for dashboards to update will be replaced by it.
Huang: “I think long term the definition of smart is someone who sits at that intersection of being technically astute, but human empathy and having the ability to infer the unspoken, around the corners, the unknowables.”
The unspoken variables are the new leverage.
The human psychology inside a market. The invisible friction in a negotiation. The instinct to build something nobody asked for yet.
You can’t spreadsheet your way there. You can’t prompt your way to that perception. It comes from decades of watching what doesn’t show up in the metrics.
Huang: “And that person might actually score horribly on the SAT.”
The future doesn’t belong to people who memorized answers.
It belongs to people who sense the questions before anyone thinks to ask.
The old system tested your ability to follow orders. The new one tests your ability to move through the unknown. And the machine can’t help you with that part.
That part is entirely on you.
Imagine the Milky Way scaled to the size of North America. How big would our Sun look?
Credit: LOVID
Over 2 million “views” now…
🤔🥲
Scientific datasets are riddled with copy-paste errors
https://www.sciencedetective.org/scientific-datasets-are-riddled-with-copy-paste-errors/ via #MarkusEnglund
An Empirical Assessment of the Cost of Dichotomization of the Outcome of Clinical Trials
By @f2harrell @stephensenn et al.
Via Statistics in Medicine
https://onlinelibrary.wiley.com/doi/10.1002/sim.70402
Amazingly thought provoking from @RuxandraTeslo : https://substack.com/home/post/p-189746201
In a better world, policymakers would constantly be asking themselves what they can learn from the unprecedented success of Operation Warp Speed.
Alas...
How is it that a paper that claims to show how to do target trial emulation does not address confounding by indication and its ramifications for data collection? https://www.jclinepi.com/article/S0895-4356(26)00080-6/fulltext?rss=yes #Statistics
Goal-Driven Flexible Bayesian Design presentation updated w/comparison of performance of frequentist group sequential designs: http://hbiostat.org/bayes/design . Frequentist approach takes far too long to make a decision by controlling something that is NOT an error prob. #Statistics #rct
#Statistics thought of the day: If you think you can find new disease subtypes by empirically clustering patients, think again: https://www.fharrell.com/post/cluster/index.html
1/2 Nail in the coffin of dichotomania: @ErikVanZwet 's paper with @stephensenn and myself just published: https://onlinelibrary.wiley.com/doi/10.1002/sim.70402 with extended discussion at https://discourse.datamethods.org/t/dichotomization #Statistics #RCT #ClinicalTrial
R to @f2harrell: 2/2 Our paper shows that despite clinical trials having to use larger sample sizes to compensate for the huge information loss from using binary outcomes, the resulting power is STILL LOWER than the power of SMALLER clinical trials using continuous outcomes. #Statistics #RCT
A 75 anni, Richard Gere ha condiviso una riflessione semplice ma profondissima sul tempo che passa.
Sono invecchiato, sì.
Ma il cuore è rimasto giovane.
L’anima non ha età: resta ferma a quando abbiamo imparato a sentire davvero.
E oggi so che questa è l’età perfetta.
Perché ogni anno lo è.
Ogni fase della vita regala qualcosa che nessun’altra può dare.
Invecchiare non è una sconfitta.
È un privilegio.
Uno che non tutti hanno la fortuna di vivere.
Non si tratta di combattere il tempo, ma di camminare al suo fianco.
Di restare vivi dentro.
Di custodire la curiosità, la dolcezza, la capacità di stupirsi.
Perché il vero segreto non è sembrare giovani.
È avere lo spirito sveglio fino all’ultimo giorno.
E quando succede, tutta la vita diventa un dono.
In 1959 BERTRAND RUSSELL - who left us 56yrs ago today - was asked what message he would most like to pass on to future generations.
His answer resonates even more strongly over 60yrs on.
Powerful video. Remember what it was like. Plot a return.
The constant stream of poor biomarker research in biomedical research journals makes me think my article "How to Do Bad Biomarker Research" must have been hugely influential :-) http://www.fharrell.com/post/badb #Statistics #biomarker
#Statistics thought of the day: Multiplicities have caused massive confusion. Rather than making arbitrary multiplicity adjustments, it is far more productive and interpretable to set the bar higher regarding actual assertions being made. https://hbiostat.org/bayes/bet/multiplicity #Clinicaltrial
Es gibt sie noch, die US-Polizisten mit 🥚🥚! 👍 😊
To see support and advocacy for Bayesian design and analysis at the highest level of @US_FDA is a thrill.
An ICE prosecutor in Dallas, James Rodden, who was previously exposed for running a white-supremacist X account praising Hitler and calling America a “White nation,” is quietly back in immigration court. After the Texas Observer revealed his online activity last year, he was pulled from court schedules and Congress demanded an investigation. Immigration and Customs Enforcement promised a review within 120 days. That deadline blew past with zero public outcome.
Now Rodden has resurfaced at the prosecutor’s table, badge on, no explanation given. ICE refuses to comment. The account is still active, just locked. Bottom line: a government lawyer who openly promotes racist ideology appears to be prosecuting immigrants again, with no transparency, no accountability, and no evidence that ICE ever did what it claimed it would do.
Same system. Same silence. Same bullshit.
https://www.texasobserver.org/ice-prosecutor-racist-account-back-at-immigration-court/