[Consistency] estimate_slopes()

[DominiqueMakowski]

Fixes #298
Fixes #288
Fixes #225

We should try to match the emmeans' output I'd say, i.e., in marginaleffects:

• Keep SE
• Keep Statistic but rename it to the actual statistic name (if possible?)

[strengejacke]

If you add

Fixes #298
Fixes #288

etc. in the beginning of the description, related issues will be closed automatically on merge.

[strengejacke]

For #301, I had to include the parameter column, which is now missing here after merging main into this PR:

trends <- datawizard::data_remove(trends, c("Statistic", "SE", "S", "CI", "df", "rowid_dedup"), verbose = FALSE) # nolint

You must check if everything is still ok in this PR, but for contrasts / contrasting slopes, we need the Parameter column, else the output makes no sense. However, sometimes that column is redundant, containing only "dx/dy". We need to check more in depth later and improve the handling.

[DominiqueMakowski]

can we add that if unique(col) == "dx/dy" we drop it?

[strengejacke]

If you add

Fixes #298
Fixes #288

etc. in the beginning of the description, related issues will be closed automatically on merge.

That's why I love the VSCode/Positron features for working with GitHub, so comfortable...

[strengejacke]

can we add that if unique(col) == "dx/dy" we drop it?

Yes, maybe we add a small helper-functions that checks for all unneeded content. Not sure if "dx/dy" will be the only exception. But for now, just checking for that value should work.

[strengejacke]

[strengejacke]

For emmeans, using estimate_means() , we do not report df and statistic.

[strengejacke]

data(mtcars)
data(iris)

dat <- mtcars
dat$gear <- as.factor(dat$gear)
dat$cyl <- as.factor(dat$cyl)
dat <<- dat

# Simple
model <- lm(vs ~ cyl, data = dat)
modelbased::estimate_means(model)
#> We selected `by = c("cyl")`.
#> Estimated Marginal Means
#> 
#> cyl |     Mean |   SE |        95% CI
#> -------------------------------------
#> 4   |     0.91 | 0.09 | [ 0.72, 1.09]
#> 6   |     0.57 | 0.11 | [ 0.34, 0.80]
#> 8   | 2.22e-16 | 0.08 | [-0.16, 0.16]
#> 
#> Marginal means estimated at cyl
modelbased::estimate_means(model, backend = "marginaleffects")
#> Estimated Marginal Means
#> 
#> cyl |     Mean |   SE |        95% CI |    t(29)
#> ------------------------------------------------
#> 4   |     0.91 | 0.09 | [ 0.72, 1.09] |    10.02
#> 6   |     0.57 | 0.11 | [ 0.34, 0.80] |     5.03
#> 8   | 2.22e-16 | 0.08 | [-0.16, 0.16] | 2.76e-15
#> 
#> Marginal means estimated at cyl

^{Created on 2025-01-09 with reprex v2.1.1}