Skip to content

reff incorrectly documented #76

@sethaxen

Description

@sethaxen

reff is documented as the relative efficiency of the unsmoothed ratios:

PSIS.jl/src/core.jl

Lines 25 to 26 in 35b9e71

- `reff`: the ratio of the effective sample size of the unsmoothed importance ratios and
the actual sample size.

PSIS.jl/src/core.jl

Lines 189 to 192 in 35b9e71

- `reff::Union{Real,AbstractArray}`: the ratio(s) of effective sample size of
`log_ratios` and the actual sample size `reff = ess/(draws * chains)`, used to account
for autocorrelation, e.g. due to Markov chain Monte Carlo. If an array, it must have the
size `(parameters...,)` to match `log_ratios`.

However, reff should be the relative efficiency of the inverse of the unsmoothed ratios. See e.g. the docstring for loo::psis : https://github.com/stan-dev/loo/blob/b94b2b148e55ece22b83ef5995e0e4b98ff1abe0/R/psis.R#L16-L26

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions