Feature Request: Contrasting Simple Slopes

[strengejacke]

Fixes #238

[DominiqueMakowski]

API-wise I think it would make sense to add a estimate_contrasts() method to apply to estimate_slopes() outputs

estimate_slopes(m) |> estimate_contrasts()

Instead of complexifying the estimate_slopes() itself

[DominiqueMakowski]

(although then the same could be said about having a working estimate_means(m) |> estimate_contrasts() workflow)

[strengejacke]

It would be the same we do with estimate_means() and estimate_contrasts() for {marginaleffects}. You just need to add the hypothesis argument to the call marginaleffects::avg_predictions(). I think we should do the same for contrasts from slopes, that avoids duplicated code. This is passed via ..., so from a user perspective, no more complexity.

(we still can think about options to pipe...)

[DominiqueMakowski]

Yes, we would still not recompute from scratch

I reckon a pipe workflow would just be syntactic sugar for:

estimate_contrasts.estimate_means <- function(x) {
  model <- extract_model(x)
  # recompute contrasts
  estimate_contrasts(model, ...)
}

[strengejacke]

Seems to work in general...

library(modelbased)
fit <- lm(Sepal.Width ~ Petal.Length * Species, data = iris)

estimate_slopes(fit, trend = "Petal.Length", backend = "marginaleffects")
#> Estimated Marginal Effects
#> 
#> Parameter    | Coefficient |       95% CI | Predicted |      p
#> --------------------------------------------------------------
#> Petal.Length |        0.33 | [0.14, 0.52] |      3.40 | < .001
#> Marginal effects estimated for Petal.Length
emmeans::emtrends(fit, specs = ~Petal.Length, var = "Petal.Length")
#> NOTE: Results may be misleading due to involvement in interactions
#>  Petal.Length Petal.Length.trend     SE  df lower.CL upper.CL
#>          3.76              0.332 0.0964 144    0.142    0.523
#> 
#> Results are averaged over the levels of: Species 
#> Confidence level used: 0.95

estimate_slopes(fit, trend = "Petal.Length", by = "Species", backend = "marginaleffects")
#> Estimated Marginal Effects
#> 
#> Parameter    |    Species | Coefficient |        95% CI |      p
#> ----------------------------------------------------------------
#> Petal.Length |     setosa |        0.39 | [-0.12, 0.90] | 0.136 
#> Petal.Length | versicolor |        0.37 | [ 0.19, 0.56] | < .001
#> Petal.Length |  virginica |        0.23 | [ 0.07, 0.39] | 0.004 
#> Marginal effects estimated for Petal.Length

estimate_contrasts(fit, contrast = "Petal.Length", by = "Species", backend = "marginaleffects")
#> Marginal Contrasts Analysis
#> 
#> Parameter              | Coefficient |        95% CI |      p
#> -------------------------------------------------------------
#> setosa - versicolor    |        0.01 | [-0.53, 0.56] | > .999
#> setosa - virginica     |        0.15 | [-0.38, 0.69] | > .999
#> versicolor - virginica |        0.14 | [-0.11, 0.39] | 0.804 
#> 
#> Marginal contrasts estimated at Petal.Length
#> p-value adjustment method: Holm (1979)

emmeans::emtrends(fit, specs = pairwise ~ Species, var = "Petal.Length")
#> $emtrends
#>  Species    Petal.Length.trend     SE  df lower.CL upper.CL
#>  setosa                  0.388 0.2600 144  -0.1264    0.902
#>  versicolor              0.374 0.0961 144   0.1843    0.564
#>  virginica               0.234 0.0819 144   0.0725    0.396
#> 
#> Confidence level used: 0.95 
#> 
#> $contrasts
#>  contrast               estimate    SE  df t.ratio p.value
#>  setosa - versicolor      0.0136 0.277 144   0.049  0.9987
#>  setosa - virginica       0.1535 0.273 144   0.563  0.8400
#>  versicolor - virginica   0.1400 0.126 144   1.108  0.5105
#> 
#> P value adjustment: tukey method for comparing a family of 3 estimates

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