An introduction to linear mixed effects models (LMMs) and their estimation. Aims to give an intuitive reason why we need LMMs followed by some theory and code that codes them (largely) from scratch.
Gaussian processes have an aura of abstract complexity - "distributions over function space". I find that linking them to linear models helps reduce the abstractness.
Some notes after reading Cranmer, Brehmer & Louppe's overview of simulation based inference.
Some notes and code after reading Baydin, Pearlmutter, Radul & Siskind's excellent paper on the magic that is automatic differentiation.
A short introduction to causal mediation analysis and the need to think carefully about potential confounding when undertaking such analyses.
Autoregressive conditional heteroscedasticity (ARCH) models have such a long name they must be great right!? I develop some ARCH models that attempt (badly) to predict Bitcoin / $US price movements.
This is a long post trying to understand structural nested mean models and their role in causal inference.