Aidan Bindoff

🦀 smoothbp has been submitted to CRAN!
Hierarchical piecewise regression with smoothed change-points in #RStats — multi-breakpoint models, spike-and-slab regularisation for automatic breakpoint selection, and a fast MCMC sampler written in Rust under the hood.
Dev version: github.com/ABindoff/smoothbp
#rstats #statistics #bayesian

If you're benchmarking your custom MCMC sampler against an industry standard (like Stan) with the help of AI, get it to compute the distance between each parameter's posterior using the 1-Wasserstein metric as well as plotting their densities. If you give the AI an objective test to pass it will be much less likely to lie about it than if you ask it to plot the densities and compare them that way.

I've been playing around with some ideas for improving reproducibility for workflows that include agentic AI/LLMs. Keen to hear other's thoughts and what you've been doing (if anything other than git)?