Cross fit generalized linear models

cross_fit_glm(
  data,
  formulas,
  cols = NULL,
  weights = NULL,
  families = gaussian(link = identity),
  fn_args = list(),
  tidy = tidy_glance,
  tidy_args = list(),
  errors = c("stop", "warn")
)

Arguments

data: A data frame
formulas: A list of formulas to apply to each subset of the data. If named, these names will be used in the model column of the output. Otherwise, the formulas will be converted to strings in the model column.
cols: Columns to subset the data. Can be any expression supported by <tidy-select>. If NULL, the data is not subset into columns. Defaults to NULL.
weights: A list of columns passed to weights in fn. If one of the elements is NULL or NA, that model will not be weighted. Defaults to NULL.
families: A list of glm model families. Defaults to gaussian("identity"), the equivalent of lm(). See family for examples.
fn_args: A list of additional arguments to glm().
tidy: A logical or function to use to tidy model output into data.frame columns. If TRUE, uses the default tidying function: tidy_glance(). If FALSE, NA, or NULL, the untidied model output will be returned in a list column named fit. An alternative function can be specified with an unquoted function name or a purrr-style lambda function with one argument (see usage with broom::tidy(conf.int = TRUE) in examples). Defaults to tidy_glance.
tidy_args: A list of additional arguments to the tidy function
errors: If "stop", the default, the function will stop and return an error if any subset produces an error. If "warn", the function will produce a warning for subsets that produce an error and return results for all subsets that do not.

Value

A tibble with a column for the model formula, columns for subsets, columns for the model family and type, columns for the weights (if applicable), and columns of tidy model output or a list column of models (if tidy = FALSE)

Examples

cross_fit_glm(
  data     = mtcars,
  formulas = list(am ~ gear, am ~ cyl),
  cols     = vs,
  families = list(gaussian("identity"), binomial("logit"))
)
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
#> # A tibble: 16 × 17
#>    model     family  link     vs term  estimate std.e…¹ statis…² p.value null.…³
#>    <chr>     <chr>   <chr> <dbl> <chr>    <dbl>   <dbl>    <dbl>   <dbl>   <dbl>
#>  1 am ~ gear gaussi… iden…     0 (Int…   -1.57  1.69e-1 -9.28e+0 7.73e-8     4  
#>  2 am ~ gear gaussi… iden…     0 gear     0.536 4.64e-2  1.15e+1 3.59e-9     4  
#>  3 am ~ gear gaussi… iden…     1 (Int…   -1.58  9.07e-1 -1.74e+0 1.08e-1     3.5
#>  4 am ~ gear gaussi… iden…     1 gear     0.538 2.33e-1  2.31e+0 3.95e-2     3.5
#>  5 am ~ gear binomi… logit     0 (Int… -177.    4.09e+5 -4.34e-4 1.00e+0    22.9
#>  6 am ~ gear binomi… logit     0 gear    50.4   1.16e+5  4.36e-4 1.00e+0    22.9
#>  7 am ~ gear binomi… logit     1 (Int…  -74.8   1.28e+4 -5.82e-3 9.95e-1    19.4
#>  8 am ~ gear binomi… logit     1 gear    18.8   3.21e+3  5.85e-3 9.95e-1    19.4
#>  9 am ~ cyl  gaussi… iden…     0 (Int…    2.54  5.65e-1  4.51e+0 3.58e-4     4  
#> 10 am ~ cyl  gaussi… iden…     0 cyl     -0.297 7.50e-2 -3.96e+0 1.12e-3     4  
#> 11 am ~ cyl  gaussi… iden…     1 (Int…    2.10  5.77e-1  3.64e+0 3.38e-3     3.5
#> 12 am ~ cyl  gaussi… iden…     1 cyl     -0.35  1.24e-1 -2.83e+0 1.52e-2     3.5
#> 13 am ~ cyl  binomi… logit     0 (Int…   79.7   1.52e+4  5.26e-3 9.96e-1    22.9
#> 14 am ~ cyl  binomi… logit     0 cyl    -10.2   1.89e+3 -5.38e-3 9.96e-1    22.9
#> 15 am ~ cyl  binomi… logit     1 (Int…   39.7   6.52e+3  6.08e-3 9.95e-1    19.4
#> 16 am ~ cyl  binomi… logit     1 cyl     -9.71  1.63e+3 -5.95e-3 9.95e-1    19.4
#> # … with 7 more variables: df.null <int>, logLik <dbl>, AIC <dbl>, BIC <dbl>,
#> #   deviance <dbl>, df.residual <int>, nobs <int>, and abbreviated variable
#> #   names ¹std.error, ²statistic, ³null.deviance

Cross fit generalized linear models

Arguments

Value

See also

Examples