CODEWITHSUNDEEP
rdataframedplyrtidyversetidyr
rnorouzian
rnorouzian

Reputation: 7517

Replacing `do()` from dplyr package with a new function with the same output

I understand the lifecycle of do() from dplyr has now been superseded.

However, in my example below, what can I use instead of do() to reproduce the exact same result that do() gives?

library(tidyverse)
library(broom)

AWD <- read.csv('https://raw.githubusercontent.com/rnorouzian/e/master/aw.csv')

AWD_long <- AWD %>%
  gather(key = "var", value = "Outcome", -Patient) %>%
 separate(var, into = c("Outcome_Measure", "Treatment"), sep = " - ") %>%
    select(Patient, Treatment, Outcome_Measure, Outcome)


AWD_long %>%
  group_by(Outcome_Measure) %>%
  do(tidy(t.test(Outcome ~ Treatment, paired = TRUE, data = .))) # what can I use to replace `do()` here?

Upvotes: 1

Views: 38

Answers (1)

akrun
akrun

Reputation: 887088

Here is an option with nest_by

library(dplyr)
library(tidyr)
library(broom)
AWD_long %>%
   nest_by(Outcome_Measure) %>%
   transmute(new = list(tidy(t.test(Outcome ~ Treatment,
          paired = TRUE, data = data))))  %>% 
   unnest(c(new))

-output

# A tibble: 3 x 9
# Groups:   Outcome_Measure [3]
#  Outcome_Measure        estimate statistic p.value parameter conf.low conf.high method        alternative
#  <chr>                     <dbl>     <dbl>   <dbl>     <dbl>    <dbl>     <dbl> <chr>         <chr>      
#1 Direct cost             6976.        4.70 0.00221         7 3464.     10487.   Paired t-test two.sided  
#2 ICU days                   1.38      1.95 0.0923          7   -0.293      3.04 Paired t-test two.sided  
#3 Length of stay (Hours)   108         5.32 0.00109         7   60.0      156.   Paired t-test two.sided

Or another option is map after doing a group_split

library(purrr)
AWD_long %>% 
   group_split(Outcome_Measure) %>% 
   map_dfr(~ tidy(t.test(Outcome ~ Treatment, paired = TRUE, data = .x)) %>% 
       bind_cols(.x['Outcome_Measure'] %>%
                     slice(1), .))

Upvotes: 1

Related Questions