Make fast_regression() function

[spsanderson]

Function:

fast_regression <- function(.data, .rec_obj, .parsnip_fns = "all",
                            .parsnip_eng = "all", .split_type = "initial_split",
                            .split_args = NULL){
  
  # Tidy Eval ----
  call <- list(.parsnip_fns) %>%
    purrr::flatten_chr()
  engine <- list(.parsnip_eng) %>%
    purrr::flatten_chr()
  
  rec_obj <- .rec_obj
  split_type <- .split_type
  split_args <- .split_args
  
  # Checks ----
  
  # Get data splits
  df <- dplyr::as_tibble(.data)
  splits_obj <- create_splits(
    .data = df, 
    .split_type = split_type,
    .split_args = split_args
  )
  
  # Generate Model Spec Tbl
  mod_spec_tbl <- fast_regression_parsnip_spec_tbl(
    .parsnip_fns = call,
    .parsnip_eng = engine
  )
  
  mod_rec_tbl <- mod_spec_tbl %>%
    dplyr::mutate(.model_recipe = list(rec_obj))
  
  mod_tbl <- mod_rec_tbl %>%
    dplyr::mutate(
      .wflw = list(
        workflows::workflow() %>%
          workflows::add_recipe(.model_recipe[[1]]) %>%
          workflows::add_model(.model_spec[[1]])
      )
    ) %>%
    dplyr::mutate(
      .fitted_wflw = list(
        parsnip::fit(.wflw[[1]], data = rsample::training(splits_obj$splits))
      )
    ) %>%
    dplyr::mutate(
      .pred_wflw = list(
        predict(.fitted_wflw[[1]], new_data = rsample::testing(splits_obj$splits))
      )
    )
  
  # Return ----
  class(mod_tbl) <- c("fst_reg_tbl", class(mod_tbl))
  attr(mod_tbl, ".parsnip_engines") <- .parsnip_eng
  attr(mod_tbl, ".parsnip_functions") <- .parsnip_fns
  attr(mod_tbl, ".split_type") <- .split_type
  attr(mod_tbl, ".split_args") <- .split_args
  return(mod_tbl)
}

Example:

> rec_obj <- recipes::recipe(mpg ~ ., data = mtcars)
> frt_tbl <- fast_regression(mtcars, rec_obj, .parsnip_eng = c("lm","glm"))
> frt_tbl
# A tibble: 3 × 8
  .parsnip_engine .parsn…¹ .pars…² .model_…³ .model…⁴ .wflw      .fitted_…⁵ .pred_…⁶
  <chr>           <chr>    <chr>   <list>    <list>   <list>     <list>     <list>  
1 lm              regress… linear… <spec[+]> <recipe> <workflow> <workflow> <tibble>
2 glm             regress… linear… <spec[+]> <recipe> <workflow> <workflow> <tibble>
3 glm             regress… poisso… <spec[+]> <recipe> <workflow> <workflow> <tibble>
# … with abbreviated variable names ¹​.parsnip_mode, ²​.parsnip_fns, ³​.model_spec,
#   ⁴​.model_recipe, ⁵​.fitted_wflw, ⁶​.pred_wflw
> class(frt_tbl)
[1] "fst_reg_tbl"      "fst_reg_spec_tbl" "tbl_df"           "tbl"             
[5] "data.frame"      
> attributes(frt_tbl)
$names
[1] ".parsnip_engine" ".parsnip_mode"   ".parsnip_fns"    ".model_spec"    
[5] ".model_recipe"   ".wflw"           ".fitted_wflw"    ".pred_wflw"     

$row.names
[1] 1 2 3

$class
[1] "fst_reg_tbl"      "fst_reg_spec_tbl" "tbl_df"           "tbl"             
[5] "data.frame"      

$.parsnip_engines
[1] "lm"  "glm"

$.parsnip_functions
[1] "all"

$.split_type
[1] "initial_split"

> 
> frt_tbl$.fitted_wflw[[1]] %>%
+   broom::glance()
# A tibble: 1 × 12
  r.squared adj.r.s…¹ sigma stati…² p.value    df logLik   AIC   BIC devia…³ df.re…⁴
      <dbl>     <dbl> <dbl>   <dbl>   <dbl> <dbl>  <dbl> <dbl> <dbl>   <dbl>   <int>
1     0.880     0.788  2.58    9.55 1.76e-4    10  -49.5  123.  137.    86.8      13
# … with 1 more variable: nobs <int>, and abbreviated variable names
#   ¹​adj.r.squared, ²​statistic, ³​deviance, ⁴​df.residual
# ℹ Use `colnames()` to see all variable names
> 
> frt_tbl$.fitted_wflw[[1]] %>%
+   broom::tidy()
# A tibble: 11 × 5
   term        estimate std.error statistic p.value
   <chr>          <dbl>     <dbl>     <dbl>   <dbl>
 1 (Intercept)  14.3      20.6       0.692    0.501
 2 cyl          -0.507     1.14     -0.443    0.665
 3 disp          0.0119    0.0252    0.473    0.644
 4 hp           -0.0279    0.0249   -1.12     0.282
 5 drat          1.72      1.89      0.912    0.379
 6 wt           -2.99      2.34     -1.28     0.223
 7 qsec          0.670     0.786     0.852    0.409
 8 vs           -0.391     2.43     -0.161    0.874
 9 am            2.27      2.22      1.02     0.326
10 gear         -0.0871    1.59     -0.0549   0.957
11 carb          0.257     0.957     0.269    0.792