The broomExtra package allows for easy augmentation of many different broom like packages in one place. It also offers the grouped_ variants of the broom functions (tidy, augment, glance). However, is this a shortcut to use purrr::map() and then the appropriate broom variant (tidy, augment, or glance)? In this case, consider tidy(). (Similar logic would be considered for the others).
library(dplyr) #mutate, group_by
#> Warning: package 'dplyr' was built under R version 4.1.2
#>
#> 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
# linear mixed effects model
broomExtra::grouped_tidy(
data = mutate(MASS::Aids2, interval = death - diag),
grouping.vars = sex,
..f = lme4::lmer,
formula = interval ~ age + (1 | status),
control = lme4::lmerControl(optimizer = "bobyqa"),
tidy.args = list(conf.int = TRUE, conf.level = 0.99)
)
#> # A tibble: 8 x 9
#> sex effect group term estimate std.error statistic conf.low conf.high
#> <fct> <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 F fixed <NA> (Inte~ 623. 161. 3.88 209. 1037.
#> 2 F fixed <NA> age -4.34 2.61 -1.66 -11.1 2.38
#> 3 F ran_pars status sd__(~ 169. NA NA NA NA
#> 4 F ran_pars Residual sd__O~ 415. NA NA NA NA
#> 5 M fixed <NA> (Inte~ 553. 62.5 8.84 392. 714.
#> 6 M fixed <NA> age -3.60 0.696 -5.17 -5.39 -1.80
#> 7 M ran_pars status sd__(~ 79.8 NA NA NA NA
#> 8 M ran_pars Residual sd__O~ 355. NA NA NA NA
#This is a shorthand to using the map function, I think:
library(tidyr)
library(purrr)
library(magrittr)
#>
#> Attaching package: 'magrittr'
#> The following object is masked from 'package:purrr':
#>
#> set_names
#> The following object is masked from 'package:tidyr':
#>
#> extract
library(broom.mixed) #also captured in broomExtra
#> Warning: package 'broom.mixed' was built under R version 4.1.2
f <- MASS::Aids2 %>%
mutate(interval = death - diag) %>%
group_by(sex) %>%
tidyr::nest()
g <- f %>%
dplyr::mutate(
fit_lmer4 = purrr::map(
data,
~ lme4::lmer(
interval ~ age + (1 | status),
data = .,
control = lme4::lmerControl(optimizer = "bobyqa")
)
),
tidy_lemr4 = purrr::map(
fit_lmer4, ~
tidy(., conf.int = TRUE, conf.level = 0.99) #from broom.mixed (also captured n broomExtar)
)
) %>%
unnest(tidy_lemr4)
g
#> # A tibble: 8 x 11
#> # Groups: sex [2]
#> sex data fit_lmer4 effect group term estimate std.error statistic conf.low
#> <fct> <lis> <list> <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 M <tib~ <lmerMod> fixed <NA> (Int~ 553. 62.5 8.84 392.
#> 2 M <tib~ <lmerMod> fixed <NA> age -3.60 0.696 -5.17 -5.39
#> 3 M <tib~ <lmerMod> ran_p~ stat~ sd__~ 79.8 NA NA NA
#> 4 M <tib~ <lmerMod> ran_p~ Resi~ sd__~ 355. NA NA NA
#> 5 F <tib~ <lmerMod> fixed <NA> (Int~ 623. 161. 3.88 209.
#> 6 F <tib~ <lmerMod> fixed <NA> age -4.34 2.61 -1.66 -11.1
#> 7 F <tib~ <lmerMod> ran_p~ stat~ sd__~ 169. NA NA NA
#> 8 F <tib~ <lmerMod> ran_p~ Resi~ sd__~ 415. NA NA NA
#> # ... with 1 more variable: conf.high <dbl>
Created on 2022-02-04 by the reprex package (v2.0.1)
The answer seems to be yes, but I want to verify that the grouped_ broom functions provide a shortcut to this syntax. (By the way, examination of the function doesn't give insight into this)
grouped_tidy
function (data, grouping.vars, ..f, ..., tidy.args = list()) {
tidy_group <- function(.x, .y) {
model <- ..f(.y = ..., data = .x)
exec(broomExtra::tidy, model, !!!tidy.args)
}
grouped_cleanup(data, enquos(grouping.vars), tidy_group)
}
