Is using the rowwise always the same as grouping by everything?

[davidhodge931]

library(tidyverse)
df <- tibble(x = runif(6), y = runif(6), z = runif(6))
# Compute the mean of x, y, z in each row

df %>% 
  rowwise() %>% 
  mutate(m = mean(c(x, y, z)))

df %>% 
  group_by(across(everything())) %>% 
  mutate(m = mean(c(x, y, z)))

[gueyenono]

Hi @davidhodge931,

Your simulated dataset does not really take advantage of the group_by() function - it does not really contain any grouping variable. So yes, in your example, they are similar. However, the results are different when you have an actual grouping variable:

library(dplyr)

d <- tibble(
  g = gl(3, 2),
  x = runif(6), 
  y = runif(6),
  z = runif(6)
)

d %>%
  group_by(g) %>%
  mutate(m = mean(c(x, y, z)))

d %>% 
  rowwise() %>% 
  mutate(m = mean(c(x, y, z)))

[davidhodge931]

Thanks @gueyenono

Having thought about this, I think that it is the same unless there is a duplicate row

library(dplyr)

d <- tibble( #new data with  last row duplicated
  g = c("A", "B", "A"), 
  x = c(1, 2, 1), 
  y = c(1, 2, 1),
  z = c(1, 2, 1)
)

d %>% 
  group_by(across(everything())) %>% 
  mutate(m = mean(x, y, z)) #breaks

d %>% 
  rowwise() %>% 
  mutate(m = mean(x, y, z)) #works