/ lab
Statistical experiments
Small pieces, each one going after a single thing about a model. Rougher than a topic page, deliberately. Two so far.
GAMLSS
Same scatter, three fits
Four datasets, same move on each: stack a line, a GAM, and a GAMLSS over the same points and see where each one runs out. A line only has a slope to work with. A GAM can bend toward the mean. A GAMLSS can also let the spread change as you move along.
explore →
Bayesian
Three problems, three posteriors
A search, a rate, a decision. Each one ends up with a distribution over the thing you didn't know, and once you have that, the questions you actually care about are just integrals over it.
explore →
source fugue