Frank Harrell

*NEW OPEN-ACCESS PAPER* “A decomposition of Fisher’s information to inform sample size for developing or updating fair and precise clinical prediction models for individual risk—part 1: binary outcomes" - published in Diagnostic & Prognostic Research rdcu.be/eveSI

#Statistics thought of the day: if you need to do power calculations for a two-group comparison on an ordinal or continuous response variable Y with or without clumping at a specific value (e.g., zero) of Y, take a look at fharrell.com/post/pop/ #rstats

🧵 7/ Unfortunately, the outcome measurement here is too flawed for this trial to change practice in my opinion. pmc.ncbi.nlm.nih.gov/articles/PMC32… And HT @f2harrell #bbr Ch. 14 also @stephensenn his figure above too

New blog article on Statistical Thinking:confidence limits for bootstrap overfitting-corrected predictive performance measures for regression models, useful for strong internal validations including confidence bands for debiased calibration curves: fharrell.com/post/bootcal #rstats

Outstanding paper! Especially suited for non-statisticians who want to get started with relaxing linearity assumptions without resorting to dreaded categorization techniques ... @Statistics

Advocates suggest administrative real-world data can substitute for #RCTs and offer propensity matching to reduce bias. But results often fail to correspond: #BMJ study showed direction of effect differed in 31% & CI failed to include #RCT estimate in 56% pubmed.ncbi.nlm.nih.gov/26858277/

#Statistics thought of the day: hbiostat.org/bbr/htest.html…

1/5 How reproducible are phase III oncology trials? 📊 With @AlexSherryMD @ErikVanZwet @ebludmir et al. We analyzed 632 RCTs (~496,000 pts). The data offer both reassurance and caution for clinicians and researchers. pubmed.ncbi.nlm.nih.gov/40628175/ 👇#MsaouelLab

“Conservative commitments to past usage” (I love the phrase) hinders progress in #Statistics in a number of areas.

The author seems to be making the age-old error of using quantile groups to understand risk patterns. If the type of food consumed matters it will matter on an individual dose-response basis, not according to how many people are like you. sensible-med.com/p/the-60-year-…

#Statistics thought of the day: general adjusted pseudo R^2 measures applying to a wide variety of statistical models and types of Y: hbiostat.org/bib/r2

My online Quarto-based html books and blog use hypothes.is to allow readers to add Q&A, comments, and annotations. I've added some tips for adding drawings and diagrams to annotations here: hbiostat.org/comment #Statistics #rmscourse #bbr

#Statistics thought of the day: challenges and resources for analysis of high-dimensional data: hbiostat.org/hdata

Another reason to embrace longitudinal ordinal models! onlinelibrary.wiley.com/doi/10.1002/si…

Very interesting paper on causal inference from statistical models; perhaps more actionable than most causal inference papers: tandfonline.com/doi/full/10.10… #Statistics

Today we start the 4-day Regression Modeling Strategies course. Always energized by the bright &motivated participants. Partnering a PhD biostatistician (me) w/a PhD epidemiologist (Drew Levy) creates a nice combination of viewpoints. #rmscourse hbiostat.org/rms

Just reading a 1984 paper on Cox regression models that casually drops terms like "general index of predictive discrimination" and "incomplete principal components regression" which I'm trying to understand for thesis!!! These guys were cooking before overfitting was even a thing

Registration for tomorrow's 1-day course on R, the rms package, and linear models ends at 8pm CDT today. There's still room for those interested. hbiostat.org/doc/rms/1day #rstats #statistics