When 6+2 isn't the right tool: an introduction to Bayesian Logistic Regression Models (BLRM) for dose-toxicity modeling and the dose-optimization thinking behind the FDA's Project Optimus, with a basic R/Stan implementation.
A walkthrough of the 6+2 active-plus-placebo cohort design common in non-oncology, non-vaccine first-in-human trials, with R code simulating a cohort safety review.
What a statistician actually does in an early-phase clinical trial — and why the hardest part happens long before the data arrives. Part 1 of a series.
Announcing the CRAN release of bnns 1.0.0, an R package for Bayesian neural networks powered by Stan.