R: Minimum and maximum values by row with unknown number of columns

user_9380092 0
Viewed 1980

For a data frame, I need to find the minimum and maximum values, by row, for an unknown number of columns from column 2 onwards. Here is an example: 

library(tidyverse)

# test data
(test_data <- tibble(id = c(1:9), 
                     x = runif(9), 
                     x2 = runif(9),
                     x3 = runif(9)))
samples = 100    

# This example, which specifies the column names, correctly finds the min and max values by row
(test_1 <- test_data %>% 
  rowwise() %>%
  mutate(min_val = min(x, x2, x3), max_val = max(x, x2, x3)))

# This example does not
(test_2 <- test_data %>% 
    rowwise() %>%
    mutate(min_val = min(x:x3), max_val = max(x:x3)))

What I actually want to do is something like 

mutate(min_val = min([,2:samples+1]), max_val = max([,2:samples+1])))

because (1) I wish to retain the id column (for later joining with another data frame), and (2) specifying by column position seems an obvious way of doing this as I am unconcerned with the column name and samples may be large.

Thank you!

Edited example

This (as suggested)

test_data %>%
  nest(-id) %>%                         # nest rest of columns apart from id
  mutate(min_val = map(data, min),      # get min and max
         max_val = map(data, max)) %>%
  unnest()

works on the original test data. However, the real-world data has repetitions of id, e.g.

(test_data <- tibble(id = c(1:9, 1:9), 
                     x = runif(18), 
                     x2 = runif(18),
                     x3 = runif(18)))

and this results in 'Error: All nested columns must have the same number of elements.'.

dplyr r
2 Answers

A possible tidyverse solution is to nest any columns you have, apart from id, and then use map to get min and max. You won't need to specify any column names:

library(tidyverse)

# test data
(test_data <- tibble(id = c(1:9), 
                     x = runif(9), 
                     x2 = runif(9),
                     x3 = runif(9)))
samples = 100    

test_data %>%
  nest(-id) %>%                         # nest rest of columns apart from id
  mutate(min_val = map(data, min),      # get min and max
         max_val = map(data, max)) %>%
  unnest()                              # unnest columns

# # A tibble: 9 x 6
#      id min_val max_val      x     x2    x3
#   <int>   <dbl>   <dbl>  <dbl>  <dbl> <dbl>
# 1     1  0.0217   0.239 0.130  0.0217 0.239
# 2     2  0.125    0.814 0.625  0.814  0.125
# 3     3  0.281    0.770 0.331  0.770  0.281
# 4     4  0.123    0.868 0.123  0.644  0.868
# 5     5  0.149    0.340 0.149  0.340  0.337
# 6     6  0.496    0.865 0.596  0.865  0.496
# 7     7  0.0766   0.984 0.0766 0.656  0.984
# 8     8  0.272    0.926 0.702  0.926  0.272
# 9     9  0.433    0.912 0.912  0.433  0.590

In the case of having multiple ids you can use this:

test_data %>%
  mutate(row_id = row_number()) %>%     # create a row identifier
  nest(-id, -row_id) %>%                # nest rest of columns apart from id and row id
  mutate(min_val = map(data, min),      # get min and max
         max_val = map(data, max)) %>%
  unnest()                              # unnest columns
user_51934850

Here is one option with pmin/pmax 

library(tidyverse)
test_data %>% 
     mutate(min_val = pmin(!!! rlang::syms(names(.)[-1])),
            max_val = pmax(!!! rlang::syms(names(.)[-1])))
# A tibble: 9 x 6
#     id     x     x2     x3 min_val max_val
#  <int> <dbl>  <dbl>  <dbl>   <dbl>   <dbl>
#1     1 0.293 0.255  0.501   0.255    0.501
#2     2 0.225 0.605  0.139   0.139    0.605
#3     3 0.704 0.371  0.0939  0.0939   0.704
#4     4 0.519 0.672  0.552   0.519    0.672
#5     5 0.663 0.673  0.725   0.663    0.725
#6     6 0.920 0.320  0.138   0.138    0.920
#7     7 0.280 0.904  0.223   0.223    0.904
#8     8 0.764 0.198  0.688   0.198    0.764
#9     9 0.802 0.0442 0.0765  0.0442   0.802

data

set.seed(24)
test_data <- tibble(id = c(1:9), 
                    x = runif(9), 
                    x2 = runif(9),
                    x3 = runif(9))
user_37322710
Related