No server sampled this and no Python ran to sample it — the page holds a wasm module built from a PyMC log density and a NUTS sampler, and nothing else. Press sample again and the chains run from scratch.
Parameters are in the space the sampler works in, so a name ending
_log__ is on the log scale. The dashed red line is where
nutpie put the mean for the same model.
R-hat, ESS and the rank plots from arviz-stats itself, running on Pyodide in a worker. The first press downloads about 24 MB; after that every new sample is diagnosed as it lands.