Reference / binference.diagnostics.ess

binference.diagnostics.ess¶

binference.diagnostics.ess(x, kind='bulk')[source]¶

Effective sample size of a scalar quantity.

Parameters:

xarray_like, shape (n_draws,) or (n_chains, n_draws)
kind{“bulk”, “tail”, “mean”}: Bulk, tail or mean ESS. "bulk": rank-normalised split-chain ESS; "tail": minimum of the ESS of the 5% and 95% quantile indicators; "mean": ESS of the raw (split) draws, appropriate for the Monte Carlo error of the mean.