Interesting blogs and packages
Here is a short list of blogs and packages that I keep coming back to when working on Bayesian models, R-INLA workflows, and applied statistics.
I will add to this list over time whenever I run into references that are both practically useful and technically solid.
Blogs
- YoungStatS: short posts on modern statistical ideas, often with clear entry points into current research.
- Andrew Gelman’s blog: broad coverage of Bayesian modelling, applied inference, model criticism, and statistical practice.
- From the Bottom of the Heap: Gavin Simpson’s excellent blog on smoothing, GAMs, and practical modelling issues in
R. - Finn Lindgren’s website: talks and material around
INLA, latent Gaussian models, spatial statistics, and related methodology. - Tristan Mahr’s blog: consistently useful posts on Bayesian workflows, mixed models, and effective communication of model results.
- Post-Bayes seminar series: a valuable collection of talks and reading material around generalized Bayes, robust updating, and post-Bayesian statistical methods.
R Packages
INLA: fast approximate Bayesian inference for latent Gaussian models.inlabru: a very convenient interface aroundINLA, especially for structured model specification.brms: a flexible formula interface to Bayesian models backed by Stan.mgcv: still one of the best references for smooth regression and penalised spline modelling.
Python Packages
JAX: a high-performance numerical computing library for Python with automatic differentiation, JIT compilation, and a NumPy-like API.GPJax: a Gaussian process library built on top of JAX, with a clean interface for kernels, posterior prediction, and scalable GP modelling.PyMC: a mature probabilistic programming framework for Bayesian modelling, with strong support for model specification, posterior analysis, and diagnostics.BlackJAX: a JAX-native library of Bayesian inference algorithms, especially useful when you want direct access to modern MCMC and variational building blocks.NumPyro: a probabilistic programming library built on JAX, particularly attractive for fast Hamiltonian Monte Carlo workflows and flexible Bayesian modelling.