Dplyr::last - rows dropped if one variable can't be computed in summarise-CodePudding

I have a data frame I want to summarise, for some of the groups, some variables should return NA, but instead the whole row is removed. Toy example Df:

df=data.frame(button=c(1,2,3,3,3,2),group=c(1,1,1,2,2,2),RT=c(100,110,120,130,140,150))

When I summarise without using "last" I get as expected:

df%>%dplyr::group_by(group) %>%dplyr::summarize(RT=mean(RT), RT.button1=mean(RT[button==1]))
# A tibble: 2 x 3
  group    RT RT.button1
* <dbl> <dbl>      <dbl>
1     1   110        110
2     2   140        NaN

But when I use last, instead the row is removed

df%>%dplyr::group_by(group) %>%dplyr::summarize(RT=mean(RT), RT.button1=mean(RT[button==1]),RT.last.button1=last(RT[button==1]))
# A tibble: 1 x 4
# Groups:   group [1]
  group    RT RT.button1 RT.last.button1
  <dbl> <dbl>      <dbl>           <dbl>
1     1   110        110             110

Is there any way to get "last" to return NA instead of removing the row? I'd be very grateful for any pointers!

CodePudding user response：

This is certainly because you are using data.table::last instead of dplyr::last.

With data.table::last:

df %>% 
  dplyr::group_by(group) %>% 
  dplyr::summarize(RT = mean(RT), 
                   RT.button1 = mean(RT[button == 1]),
                   RT.last.button1 = data.table::last(RT[button == 1]))

# Groups:   group [1]
  group    RT RT.button1 RT.last.button1
  <dbl> <dbl>      <dbl>           <dbl>
1     1   110        110             110

With dplyr::last:

df %>% 
  dplyr::group_by(group) %>% 
  dplyr::summarize(RT = mean(RT), 
                   RT.button1 = mean(RT[button == 1]),
                   RT.last.button1 = dplyr::last(RT[button == 1]))
# A tibble: 2 x 4
  group    RT RT.button1 RT.last.button1
  <dbl> <dbl>      <dbl>           <dbl>
1     1   110        110             110
2     2   140        NaN              NA