Add step_adjust_linear() for linear adjustment

[mattansb]

This PR addresses the feature request in #275.

I have named the step step_adjust_linear() - its functionality can be seen in the original issue.

Here is the example from the docs:

library(embed)
#> Loading required package: recipes
#> Loading required package: dplyr
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
#> 
#> Attaching package: 'recipes'
#> The following object is masked from 'package:stats':
#> 
#>     step
library(ggplot2)

data("penguins", package = "modeldata")
penguins <- na.omit(penguins)

p <- ggplot(penguins, aes(flipper_length_mm, body_mass_g, color = sex)) +
  geom_point(aes(shape = species)) +
  stat_ellipse() +
  labs(title = "No adjustment")

p

recipe <- recipe(body_mass_g ~ ., data = penguins) |>
  step_adjust_linear(
    flipper_length_mm,
    body_mass_g,
    remove_vars = vars(species),
    keep_vars = vars(sex),
    drop = "none" # keep all variables in the baked data
  )

p +
  (prep(recipe) |>
    bake(new_data = penguins)) +
  labs(title = "Adjustment for species")

^{Created on 2026-09-23 with reprex v2.1.1}

---

Since the step does not depend on any package other than stats I'm not sure if {embed} or {recipes} are the right place for it.

[EmilHvitfeldt]

I think that {embed} is the right place for this. i'll try to look at this later this week!

[mattansb]

This step can also be used for causal inference - and since it can be used with any other previous steps, it can also adjust non-linear terms. Here's a very simple example:

library(embed)
#> Loading required package: recipes
#> Loading required package: dplyr
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
#> 
#> Attaching package: 'recipes'
#> The following object is masked from 'package:stats':
#> 
#>     step

n <- 500

set.seed(333)

dat <- tibble(
  Z = rnorm(n),
  X = 3^Z + 1.3 * Z^2 - rnorm(n),
  Y = 30 * Z - 4 * X + 100 * rnorm(n)
)

plot(dat)

rec <- recipe(Y ~ ., data = dat) |>
  step_spline_b(Z, deg_free = 5) |>
  step_adjust_linear(X, remove_vars = vars(starts_with("Z")))

baked <- prep(rec) |> bake(new_data = dat, stop_at = 2)

lm(Y ~ X, data = dat) # confounded
#> 
#> Call:
#> lm(formula = Y ~ X, data = dat)
#> 
#> Coefficients:
#> (Intercept)            X  
#>      -6.202       -0.587

lm(Y ~ X, data = baked) # should be about ~4
#> 
#> Call:
#> lm(formula = Y ~ X, data = baked)
#> 
#> Coefficients:
#> (Intercept)            X  
#>       2.070       -3.199

^{Created on 2026-09-23 with reprex v2.1.1}