[Feature] Add report.compare.loo

[DominiqueMakowski]

> x <- brms::loo_compare(brms::add_criterion(m1, "loo"),
+                        brms::add_criterion(m2, "loo"),
+                        model_names=c("m1", "m2"))
> report(x)

The difference in predictive accuracy, as index by Expected Log Predictive Density (ELPD), suggests that 'm2' is the best model, followed by 'm1' (diff =
-9.70, Z-diff = -2.98)

[mattansb]

Made some changes to report if LOO or WAIC was used and to allow for ELPD or IC to be reported, and reporting the effective number of parameters (p).

library(loo)
library(rstanarm)
library(report)

m1 <- stan_glm(mpg ~ 1, data = mtcars, refresh = 0)
m2 <- stan_glm(mpg ~ am, data = mtcars, refresh = 0)
m3 <- stan_glm(mpg ~ hp, data = mtcars, refresh = 0)

c.loo <- loo_compare(loo(m1), loo(m2), loo(m3))
c.waic <- loo_compare(waic(m1), waic(m2), waic(m3))

report(c.loo)
#> The difference in predictive accuracy, as index by Expected Log Predictive
#> Density (ELPD-LOO), suggests that 'm3' is the best model (effective number of
#> parameters (ENP) = 3.66), followed by 'm2' (diff = -7.05, ENP = 2.82, z-diff =
#> -2.39) and 'm1' (diff = -12.99, ENP = 1.80, z-diff = -3.77).
report(c.loo, index = "IC")
#> The difference in predictive accuracy, as index by Leave-One-Out CV Information
#> Criterion (LOOIC), suggests that 'm3' is the best model (effective number of
#> parameters (ENP) = 3.66), followed by 'm2' (diff = 14.10, ENP = 2.82, z-diff =
#> 2.39) and 'm1' (diff = 25.99, ENP = 1.80, z-diff = 3.77).

report(c.waic)
#> The difference in predictive accuracy, as index by Expected Log Predictive
#> Density (ELPD-WAIC), suggests that 'm3' is the best model (effective number of
#> parameters (ENP) = 3.52), followed by 'm2' (diff = -7.15, ENP = 2.78, z-diff =
#> -2.45) and 'm1' (diff = -13.12, ENP = 1.78, z-diff = -3.90).
report(c.waic, index = "IC")
#> The difference in predictive accuracy, as index by Widely Applicable
#> Information Criterion (WAIC), suggests that 'm3' is the best model (effective
#> number of parameters (ENP) = 3.52), followed by 'm2' (diff = 14.29, ENP = 2.78,
#> z-diff = 2.45) and 'm1' (diff = 26.24, ENP = 1.78, z-diff = 3.90).

^{Created on 2024-03-26 with reprex v2.1.0}

[DominiqueMakowski]

very cool! will be useful for teaching

[DominiqueMakowski]

@rempsyc I think this is good to be merged

[rempsyc]

Thanks, but we see:

── R CMD check results ───────────────────────────────────── report 0.5.8.1 ────
Duration: 5m 4.3s

❯ checking Rd cross-references ... WARNING
  Missing link or links in documentation object 'report.compare.loo.Rd':
    ‘loo_compare’
  
  See section 'Cross-references' in the 'Writing R Extensions' manual.

❯ checking Rd \usage sections ... WARNING
  Undocumented arguments in documentation object 'report.compare.loo'
    ‘...’
  
  Functions with \usage entries need to have the appropriate \alias
  entries, and all their arguments documented.
  The \usage entries must correspond to syntactically valid R code.
  See chapter ‘Writing R documentation files’ in the ‘Writing R
  Extensions’ manual.

  WORD        FOUND IN
elpd        report.compare.loo.Rd:22,23
ELPD        report.compare.loo.Rd:13
IC          report.compare.loo.Rd:13
Magnusson   report.compare.loo.Rd:23
Sivula      report.compare.loo.Rd:23
Vehtari     report.compare.loo.Rd:23
Error:
! Spelling errors found. Either correct them or update `inst/WORDLIST`
  using `spelling::update_wordlist()`.

── Building function reference ─────────────────────────────────────────────────
Error in `build_reference_index()`:
! All topics must be included in reference index
✖ Missing topics: report.compare.loo
ℹ Either add to _pkgdown.yml or use @keywords internal

[DominiqueMakowski]

should be fixed

[rempsyc]

Thanks, but we still see:

  WORD        FOUND IN
elpd        report.compare.loo.Rd:22,23
ELPD        report.compare.loo.Rd:13
IC          report.compare.loo.Rd:13
Magnusson   report.compare.loo.Rd:23
Sivula      report.compare.loo.Rd:23
Vehtari     report.compare.loo.Rd:23
Error:
! Spelling errors found. Either correct them or update `inst/WORDLIST`
  using `spelling::update_wordlist()`.

── Building function reference ─────────────────────────────────────────────────
Error in `build_reference_index()`:
! All topics must be included in reference index
✖ Missing topics: report.compare.loo
ℹ Either add to _pkgdown.yml or use @keywords internal

Some files don't follow the style guide. Run styler::style_pkg() with latest {styler}.

Error: Process completed with exit code 1.

[DominiqueMakowski]

[rempsyc]

Your wishes are my commands. Why haven't you started with this, Master Sidious? Let me get to work and fix all those remaining issues now

[rempsyc]

• Error in build_reference_index() ✅
• Some files don't follow the style guide ✅
• Lints ✅

Snapshots are still failing, but I'm afraid we'll have to fix them in a different PR :/

[rempsyc]

Can you update the readme NEWS.md with the addition of this new function? I don't know what it does 😬

[DominiqueMakowski]

Can you update the readme

No need to put it in the README I'd say, it's quite a niche and mostly "convenience" feature (a tweet about it will suffice :)

[rempsyc]

Boy am I tired hehe, I meant the NEWS.md 😂 It is done now, I have updated it, sorry about that

[strengejacke]

@rempsyc waiting for Dom to add tests and fix check issues...