Plotting PSIS results

PSIS.jl includes plotting recipes for PSISResult using any Plots.jl backend, as well as the utility plotting function PSISPlots.paretoshapeplot.

We demonstrate this with a simple example.

using PSIS, Distributions
proposal = Normal()
target = TDist(7)
ndraws, nchains, nparams = (1_000, 1, 20)
x = rand(proposal, ndraws, nchains, nparams)
log_ratios = logpdf.(target, x) .- logpdf.(proposal, x)
result = psis(log_ratios)

PSISResult with 1000 draws, 1 chains, and 20 parameters
Pareto shape (k) diagnostic values:
                    Count      Min. ESS
 (-Inf, 0.5]  good  4 (20.0%)  959
  (0.5, 0.7]  okay  9 (45.0%)  938
    (0.7, 1]  bad   7 (35.0%)  ——

Plots.jl

PSISResult objects can be plotted directly:

using Plots
plot(result; showlines=true, marker=:+, legend=false, linewidth=2)

This is equivalent to calling PSISPlots.paretoshapeplot(result; kwargs...).