Jared Huling

@jaredhuling.bsky.social

Associate Professor of Biostatistics, University of Minnesota https://jaredhuling.org/

With the latest update on CRAN, I've made additional changes to fastglm, including Firth bias correction for all standard GLMs. If you need to fit negative binomial models, try fastglm, as it's much more likely to converge than MASS::glm.nb(), see cran.r-project.org/web/packages...

Negative Binomial Stability Benchmark: fastglm vs MASS

cran.r-project.org

Jared Huling@jaredhuling.bsky.social · 3mo ago

I've made some big changes to the fastglm package with version 0.1.0 on cran: cran.r-project.org/package=fast... - updates that further improve speed - three new models for more complex outcome (neg binom, hurdle, zero infl) - First bias correction - big data support ** Details outlined below:

This is huge. You could have perfectly good cholesterol, do all the right things in terms of diet, exercise and sleep and yet still be at major risk of CVD due to high LP(a). Lots of promising therapies and we'll get to hear whether one of them works this year

John Mullahy@johnmullahy.bsky.social · 2mo ago

"[Novartis, Amgen and Lilly are] developing medicines that lower the level of lipoprotein(a), a fatty particle in the blood that is known as Lp(a)... Tens of millions of people in the U.S. have elevated Lp(a), an inherited and largely hidden driver of heart disease." www.wsj.com/health/pharm...

I've made some big changes to the fastglm package with version 0.1.0 on cran: cran.r-project.org/package=fast... - updates that further improve speed - three new models for more complex outcome (neg binom, hurdle, zero infl) - First bias correction - big data support ** Details outlined below:

fastglm: Fast and Stable Fitting of Generalized Linear Models using 'RcppEigen'

Fits generalized linear models efficiently using 'RcppEigen'. The iteratively reweighted least squares implementation utilizes the step-halving approach of Marschner (2011) &lt;<a href="https://doi.or...

cran.r-project.org

The City of Ottawa posted the raw speed data from before and after Doug Ford banned speed cameras and the effect is so obvious it's not even necessary put a line marking when that happened. This chart shows the percentage of drivers going 15km/h or more over the posted limit.

Bild

The Canvas hack is "the biggest student data privacy disaster in history." For years critics have been warning about the centralization of EdTech. Well, millions of students being locked out is what happens when you do exactly that www.404media.co/the-biggest-...

'The Biggest Student Data Privacy Disaster in History': Canvas Hack Shows the Danger of Centralized EdTech

Messages could include "medical circumstances, accessibility accommodations, disputes, sexual assault allegations," and more.

404media.co

I've made some big changes to the fastglm package with version 0.1.0 on cran: cran.r-project.org/package=fast... - updates that further improve speed - three new models for more complex outcome (neg binom, hurdle, zero infl) - First bias correction - big data support ** Details outlined below:

fastglm: Fast and Stable Fitting of Generalized Linear Models using 'RcppEigen'

Fits generalized linear models efficiently using 'RcppEigen'. The iteratively reweighted least squares implementation utilizes the step-halving approach of Marschner (2011) &lt;<a href="https://doi.or...

cran.r-project.org

I've now in the past two weeks received multiple paper reviews for papers I've submitted that were very obviously written by AI. As a field I think we need to re-assess the role of editors and associate editors in the review process

Although AI improves performance during assisted sessions, performance drops sharply once it's removed. And relative to controls, AI-assisted participants also gave up more frequently on test problems.

Bild

🚨New preprint and our results are rather concerning.. We find the "boiling frog" equivalent of AI use. Using large-scale RCTs, we provide *casual* evidence that AI assistance reduces persistence and hurts independent performance. And these effects emerge after just 10–15 minutes of AI use! 1/

Bild