add butcher methods for survival::coxph()

[simonpcouch]

Related to #234. Also no NEWS update.🐻

There's a $terms slot that carries around a hefty environment that we'd usually sub out with rlang::base_env(), but the predict.coxph() source taps into it quite a bit.

library(survival)
library(butcher)

example_data <-
  tibble::tibble(
    time = rpois(1000, 2) + 1,
    status = rbinom(1000, 1, .5),
    x = rpois(1000, .5),
    covar = rbinom(1000, 1, .5)
  )

example_data
#> # A tibble: 1,000 × 4
#>     time status     x covar
#>    <dbl>  <int> <int> <int>
#>  1     1      0     0     0
#>  2     2      0     0     0
#>  3     2      1     0     1
#>  4     3      0     1     0
#>  5     5      0     1     1
#>  6     1      0     0     1
#>  7     2      0     0     1
#>  8     7      1     0     1
#>  9     2      1     2     1
#> 10     2      0     1     1
#> # … with 990 more rows

make_big_model <- function() {
  boop <- runif(1e6)
  coxph(Surv(time, status) ~ x + strata(covar), example_data)
}

res <- make_big_model()

weigh(res)
#> # A tibble: 20 × 2
#>    object                    size
#>    <chr>                    <dbl>
#>  1 terms                 8.01    
#>  2 formula               8.00    
#>  3 residuals             0.0722  
#>  4 y                     0.0177  
#>  5 linear.predictors     0.00805 
#>  6 call                  0.00146 
#>  7 concordance           0.000752
#>  8 coefficients          0.00028 
#>  9 means                 0.00028 
#> 10 wald.test             0.00028 
#> 11 var                   0.000224
#> 12 xlevels.strata(covar) 0.000176
#> 13 method                0.000112
#> 14 loglik                0.000064
#> 15 score                 0.000056
#> 16 iter                  0.000056
#> 17 n                     0.000056
#> 18 nevent                0.000056
#> 19 assign.x              0.000056
#> 20 timefix               0.000056
weigh(butcher(res))
#> # A tibble: 20 × 2
#>    object                    size
#>    <chr>                    <dbl>
#>  1 terms                 8.01    
#>  2 residuals             0.0722  
#>  3 linear.predictors     0.00805 
#>  4 formula               0.00149 
#>  5 call                  0.00146 
#>  6 concordance           0.000752
#>  7 coefficients          0.00028 
#>  8 means                 0.00028 
#>  9 wald.test             0.00028 
#> 10 var                   0.000224
#> 11 xlevels.strata(covar) 0.000176
#> 12 method                0.000112
#> 13 loglik                0.000064
#> 14 score                 0.000056
#> 15 iter                  0.000056
#> 16 n                     0.000056
#> 17 nevent                0.000056
#> 18 assign.x              0.000056
#> 19 timefix               0.000056
#> 20 y                     0.000048

# try to make those terms more minimal....
attr(res$terms, ".Environment") <- rlang::base_env()

# but noooo!
predict(res, example_data[1:100,])
#> Error in is.data.frame(data): object 'example_data' not found

^{Created on 2022-12-22 with reprex v2.0.2}

[juliasilge]

Looks good! 🚀

[github-actions]

This pull request 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.