options are ignored in step_ns

Possible bug as the options argument is not returning the same output as splines::ns.

I'd like to specify a spline for a continuous variable using step_ns. The docs say I can pass arguments k (knot locations) and Boundary.knots via options argument to step_ns.

However, doing so does not return the same number of basis vectors that I would get were I to use model.matrix and spline::ns. See below

library(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

d <- tibble(x = seq(-2, 2, 0.01))

recipe(~., data=d) %>% 
  step_ns(x, options = list(k = seq(-1, 1, 0.125), Boundary.knots = c(-1.5, 1.5))) %>% 
  prep %>% 
  bake(new_data=d) %>% 
  ncol
#> [1] 2


model.matrix(~splines::ns(x, k = seq(-1, 1, 0.125), Boundary.knots = c(-1.5, 1.5)), data=d) %>% 
  ncol 
#> [1] 19

^{Created on 2024-04-12 with reprex v2.0.2}

@EmilHvitfeldt is this an actual bug? If so I'd like to know so I can take a different approach on a problem

[EmilHvitfeldt]

I'm pretty sure this is a bug, or at least an unintuitive interface. It appears that the deg_free argument is taking over,

https://github.com/tidymodels/recipes/blob/main/R/ns.R#L106-L130

If you need things working now, I would suggest you use step_spline_natural()

cross-posted from https://stackoverflow.com/questions/78317941/step-ns-is-not-returning-expected-number-of-basis-functions

Ouch. Ok, will try another approach. @EmilHvitfeldt happy to contribute if you cc me on the ticket

[EmilHvitfeldt]

a fix is here with #1338.

A related issue is that you are using partial matching of knots by using the argument name k. This is not supported in options argument for recipes steps, and a plan to fix that issue is listed here #1269

library(recipes)

d <- tibble(x = seq(-2, 2, 0.01))

recipe(~., data = d) %>% 
  step_ns(x, options = list(knots = seq(-1, 1, 0.125), 
                            Boundary.knots = c(-1.5, 1.5))) %>% 
  prep() %>% 
  bake(new_data = d) %>%
  dim()
#> [1] 401  18

# -1 in `model.matrix()` to avoid intercept
model.matrix(~splines::ns(x, 
                          knots = seq(-1, 1, 0.125), 
                          Boundary.knots = c(-1.5, 1.5)) - 1, data = d) %>%
  dim()
#> [1] 401  18

^{Created on 2024-06-07 with reprex v2.1.0}

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex https://reprex.tidyverse.org) and link to this issue.