dplyr summarise_each with na.rm

Question

Is there a way to instruct dplyr to use summarise_each with na.rm=TRUE? I would like to take the mean of variables with summarise_each("mean") but I don't know how to specify it to ignore missing values.

Answer 1

Following the links in the doc, it seems you can use funs(mean(., na.rm = TRUE)):

library(dplyr)
by_species <- iris %>% group_by(Species)
by_species %>% summarise_each(funs(mean(., na.rm = TRUE)))

Answer 2

update

the current dplyr version strongly suggests the use of across instead of the more specified functions summarise_all etc.

Translating the below syntax (naming the functions in a named list) into across could look like this:

library(dplyr)
ggplot2::msleep %>% 
  select(vore, sleep_total, sleep_rem) %>%
  group_by(vore) %>%
  summarise(across(everything(), .f = list(mean = mean, max = max, sd = sd), na.rm = TRUE))

#> # A tibble: 5 x 7
#>   vore  sleep_total_mean sleep_total_max sleep_total_sd sleep_rem_mean
#>   <chr>            <dbl>           <dbl>          <dbl>          <dbl>
#> 1 carni            10.4             19.4           4.67           2.29
#> 2 herbi             9.51            16.6           4.88           1.37
#> 3 inse~            14.9             19.9           5.92           3.52
#> 4 omni             10.9             18             2.95           1.96
#> 5 <NA>             10.2             13.7           3.00           1.88
#> # ... with 2 more variables: sleep_rem_max <dbl>, sleep_rem_sd <dbl>

---

older answer

summarise_each is deprecated now, here an option with summarise_all.

• One can still specify na.rm = TRUE within the funs argument (cf @flodel 's answer: just replace summarise_each with summarise_all ).
• But you can also add na.rm = TRUE after the funs argument.

That is useful when you want to call more than only one function, e.g.:

edit

the funs() argument is now (soft)deprecated, thanks to comment @Mikko. One can use the suggestions that are given by the warning, see below in the code. na.rm can still be specified as additional argument within summarise_all.

I used ggplot2::msleep because it contains NAs and shows this better.

library(dplyr)

ggplot2::msleep %>% 
  select(vore, sleep_total, sleep_rem) %>%
  group_by(vore) %>%
  summarise_all(funs(mean, max, sd), na.rm = TRUE)
#> Warning: funs() is soft deprecated as of dplyr 0.8.0
#> Please use a list of either functions or lambdas: 
#> 
#>   # Simple named list: 
#>   list(mean = mean, median = median)
#> 
#>   # Auto named with `tibble::lst()`: 
#>   tibble::lst(mean, median)
#> 
#>   # Using lambdas
#>   list(~ mean(., trim = .2), ~ median(., na.rm = TRUE))

Answer 3

Take for instance mtcars data set

library(dplyr)

You can always use summarise to avoid long syntax:

mtcars %>%
  group_by(cyl) %>% 
  summarise(mean_mpg = mean(mpg, na.rm=T),
            sd_mpg = sd(mpg, na.rm = T))

Answer 4

I don't know if my answer will add something to the previous comments. Hopefully yes.

In my case, I had a database from an experiment with two groups (control, exp) with different levels for a specific variable (day) and I wanted to get a summary of mean and sd of another variable (weight) for each group for specific levels of the variable day.

Here is an example of my database:

animal    group           day       weight      
1.1       "control"       73        NA   
1.2       "control"       73        NA   
3.1       "control"       73        NA   
9.2       "control"       73        25.2  
9.3       "control"       73        23.4  
9.4       "control"       73        25.8   
2.1       "exp"           73        NA       
2.2       "exp"           73        NA     
10.1      "exp"           73        24.4     
10.2      "exp"           73        NA     
10.3      "exp"           73        24.6

So, for instance, in this case I wanted to get the mean and sd of the weight on day 73 for each of the groups (control, exp), omitting the NAs.

I did this with this command:

data[data$day=="73",] %>% group_by(group) %>% summarise(mean(weight[group == "exp"], na.rm=T),sd(weight[group == "exp"], na.rm=T))
data[data$day=="73",] %>% group_by(group) %>% summarise(mean(weight[group == "control"], na.rm=T),sd(weight[group == "control"], na.rm=T))

Answer 5

summarise_at function in dplyr will summarise a dataset at specific column and allow to remove NAs for each functions applied. Take iris dataset and compute mean and median for variables from Sepal.Length to Petal.Width.

library(dplyr)
summarise_at(iris,vars(Sepal.Length:Petal.Width),funs(mean,median),na.rm=T)