arviz-devs/PosteriorStats.jl: v0.3.0

PosteriorStats v0.3.0

Diff since v0.2.8

Breaking changes

• eti and hdi now return an array of IntervalSets.ClosedInterval instead of two arrays, each containing the lower and upper bounds of the interval. (#36)
• ELPD-related methods (waic/loo) when they compute standard errors now use the se_ prefix instead of the _mcse suffix. (#47)
• Changes to API and defaults in summarize. (see #63 for details)
• Stop smoothing discrete data in loo_pit. (#68)
• Make SummaryStats's indexing API non-public (#69)
• Refactor SummaryStats fields and constructor (#70)

Merged pull requests:

• Bump julia-actions/julia-downgrade-compat from 1 to 2 (#57) (@dependabot[bot])
• Pointwise log-likelihoods for non-factorizable models (#58) (@marcobonici)
• Bump Julia lower bound to v1.10 (LTS) (#59) (@sethaxen)
• Cache R dependencies in CI (#60) (@sethaxen)
• Add pointwise log-likelihoods for more observation models (#61) (@sethaxen)
• Refactor summarize (#63) (@sethaxen)
• Reorganize and simplify utilities for showing tables (#64) (@sethaxen)
• Move R-loo reference tests to their own workflow (#65) (@sethaxen)
• Remove ArviZExampleData as a test dependency (#66) (@sethaxen)
• Fix relative effiency computation in loo docstring (#67) (@sethaxen)
• Stop smoothing discrete data in loo_pit (#68) (@sethaxen)
• Remove SummaryStats's fields, type parameters, and indexing interface from the public API (#69) (@sethaxen)
• Refactor SummaryStats (#70) (@sethaxen)

Closed issues:

• Changing default CI probability (#35)
• LOO and ELPD for non factorizable models (#49)
• Calculation of relative efficiency in loo docstring (#54)