warning from tidy_plus_plus for glm with subset argument

[richardjtelford]

I'm getting a warning with gtsummary::tbl_regression() for a glm with a subset argument (initial bug report ddsjoberg/gtsummary/issues/2111)

The problem is coming from broom.helpers::tidy_plus_plus()

mod <- glm(mpg ~ gear, data = mtcars, subset = mpg < 30)
broom.helpers::tidy_plus_plus(mod)
#> Warning in tcm * w: longer object length is not a multiple of shorter object
#> length
#> # A tibble: 1 × 17
#>   term  variable var_label var_class var_type   var_nlevels contrasts
#>   <chr> <chr>    <chr>     <chr>     <chr>            <int> <chr>    
#> 1 gear  gear     gear      numeric   continuous          NA <NA>     
#> # ℹ 10 more variables: contrasts_type <chr>, reference_row <lgl>, label <chr>,
#> #   n_obs <dbl>, estimate <dbl>, std.error <dbl>, statistic <dbl>,
#> #   p.value <dbl>, conf.low <dbl>, conf.high <dbl>

^{Created on 2024-12-23 with reprex v2.1.1}

The underlying issue is that in broom.helpers:::model_get_n.glm()

` `` r
tcm <- model_compute_terms_contributions(model)
if (is.null(tcm)) {
return(NULL)
}
w <- model_get_weights(model)

the number of rows in `tcm` from `model_compute_terms_contributions()` reflects the original data.frame, whereas `w` reflects the size of the data.frame after subsetting. 

``` r
nrow(mtcars) # original number of rows
#> [1] 32
nrow(mtcars[mtcars$mpg < 30, ]) # n rows after subsetting
#> [1] 28

mod <- glm(mpg ~ gear, data = mtcars, subset = mpg < 30)
broom.helpers:::model_compute_terms_contributions.default(mod) |> 
  nrow()
#> [1] 32 # original number

broom.helpers::model_get_weights(mod) |> length()
#> [1] 28 # correct


nrow(mod$model)
#> [1] 28

nrow(mod$data)
#> [1] 32

^{Created on 2024-12-23 with reprex v2.1.1}

broom.helpers:::model_compute_terms_contributions.default() is using model$data, perhaps it should be using model$model

Everything works as expected with a lm() model with the subset argument (I've not tried other models).

[larmarange]

Thanks for the report. I will have a look at it. Not sure exactly when (holiday season 😉)