Prepare CRAN submission

[strengejacke]

No description provided.

[DominiqueMakowski]

should I submit?

[strengejacke]

Let me try to fix failing tests first. Might be because we do no longer import marginaleffects from GitHub (no Remotes field), so we might skip some tests

[strengejacke]

@mattansb Do you know what's going on here, why this test suddenly fails?

    data("mtcars")
    mtcars$cyl <- factor(mtcars$cyl)
    mod <- rstanarm::stan_glm(mpg ~ cyl + hp * am, data = mtcars, refresh = 0)

    mfx <- marginaleffects::avg_slopes(mod, by = "am")
    mfx_samps <- as.data.frame(t(attr(mfx, "posterior_draws")))

    res <- hdi(mfx, ci = c(0.8, 0.9))
    expect_identical(
      as.data.frame(res[1:3]),
      data.frame(
        term = c(
          "am", "am", "am", "am", "cyl", "cyl",
          "cyl", "cyl", "cyl", "cyl", "cyl", "cyl",
          "hp", "hp", "hp", "hp"
        ),
        contrast = c(
          "1 - 0", "1 - 0", "1 - 0", "1 - 0",
          "6 - 4", "6 - 4", "8 - 4", "8 - 4",
          "6 - 4", "6 - 4", "8 - 4", "8 - 4",
          "dY/dX", "dY/dX", "dY/dX", "dY/dX"
        ),
        am = c(
          0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0,
          1, 1
        ), stringsAsFactors = FALSE
      )
    )

Failing tests:

══ Failed tests ════════════════════════════════════════════════════════════════
── Failure ('test-marginaleffects.R:36:5'): marginaleffects descrive_posterior ──
as.data.frame(res[1:3]) (`actual`) not identical to data.frame(...) (`expected`).

`names(actual)`:   "term" "contrast" "CI"
`names(expected)`: "term" "contrast" "am"

`actual$CI` is a double vector (0.8, 0.9, 0.8, 0.9, 0.8, ...)
`expected$CI` is absent

`actual$am` is absent
`expected$am` is a double vector (0, 0, 1, 1, 0, ...)
── Failure ('test-marginaleffects.R:93:5'): marginaleffects bayesfactors ───────
all(c("term", "contrast", "am") %in% colnames(outsi)) is not TRUE

`actual`:   FALSE
`expected`: TRUE 

Output hdi()

Output marginaleffects

The am column is obviously missing?

[mattansb]

Something is wrong - the am column should appear, as it does in insight::get_datagrid()

data("mtcars")
mtcars$cyl <- factor(mtcars$cyl)
mod <- rstanarm::stan_glm(mpg ~ cyl + hp * am, data = mtcars, refresh = 0)

mfx <- marginaleffects::avg_slopes(mod, by = "am")

insight::get_datagrid(mfx)
#>   term contrast am
#> 1   am    1 - 0  0
#> 2   am    1 - 0  1
#> 3  cyl    6 - 4  0
#> 4  cyl    8 - 4  0
#> 5  cyl    6 - 4  1
#> 6  cyl    8 - 4  1
#> 7   hp    dY/dX  0
#> 8   hp    dY/dX  1

When I run it, I do get all columns:

bayestestR::hdi(mfx, ci = c(0.8, 0.9))
#> Highest Density Interval
#> 
#> term | contrast |   am |        80% HDI |        90% HDI
#> --------------------------------------------------------
#> am   |    1 - 0 | 0.00 | [ 2.51,  6.27] | [ 1.93,  6.77]
#> am   |    1 - 0 | 1.00 | [ 2.63,  6.10] | [ 1.97,  6.40]
#> cyl  |    6 - 4 | 0.00 | [-5.95, -1.89] | [-6.55, -1.21]
#> cyl  |    8 - 4 | 0.00 | [-7.33,  0.00] | [-8.13,  1.29]
#> cyl  |    6 - 4 | 1.00 | [-5.95, -1.89] | [-6.55, -1.21]
#> cyl  |    8 - 4 | 1.00 | [-7.33,  0.00] | [-8.13,  1.29]
#> hp   |    dY/dX | 0.00 | [-0.08, -0.02] | [-0.08, -0.01]
#> hp   |    dY/dX | 1.00 | [-0.06, -0.02] | [-0.07, -0.02]

[mattansb]

Something is wrong here - the tested code you posted does not match the output or the failed test...

[strengejacke]

Are you using the dev-version of {marginaleffects}?

[strengejacke]

Ok, this works only with the dev version of marginaleffects, not the CRAN version. I updated the test to skip if package version is 0.24.0 or lower. If checks pass, we can submit @DominiqueMakowski

[mattansb]

Ah, I see. I thought this was already on CRAN. Oops.

[strengejacke]

@DominiqueMakowski please submit.

[strengejacke]

Remember to pull the latest changes from here ;-)
Btw, we inherit some of the docs from insight, which is not always focused on Bayesian models. You may want to change the docs for the component argument (add own description), instead of inheriting from insight.

[DominiqueMakowski]

submitted

[DominiqueMakowski]

on its way to CRAN.