R: Create multiple new columns based upon other columns

Question

Let\'s say I have a data-frame that looks like this

dd <- read.table(header = TRUE, text = \"ID week1_t week1_a  week2_t week2_a
  1      12      22       17          


        

Answer 1

From a "tidy data" perspective, your problem is that you're encoding (multiple!) pieces of data in your column names: the week number and whatever the letter stands for. I would convert to a long format where week is a column, define d = a - t, and (if necessary) convert back to wide format. But probably I'd keep it in the long format because if there are any other operations you want to do they'll probably be easier to implement on the long data (more manipulation, modeling, plotting...).

library(tidyr)
library(dplyr)

long = dd %>% 
    mutate(real_id = 1:n()) %>%
    gather(key = key, value = value, starts_with("week")) %>%
    separate(key, into = c("week", "letter")) %>% 
    spread(key = letter, value = value) %>%
    mutate(d = a - t)

head(long)
#   ID real_id  week  a  t   d
# 1  1       1 week1 22 12  10
# 2  1       1 week2  4 17 -13
# 3  1       2 week1 32 15  17
# 4  1       2 week2  5 18 -13
# 5  1       3 week1 12 24 -12
# 6  1       3 week2  6 29 -23

wide = gather(long, key = letter, value = value, a, t, d) %>%
    mutate(key = paste(week, letter, sep = "_")) %>%
    select(-week, -letter) %>%
    spread(key = key, value = value)

wide
#   ID real_id week1_a week1_d week1_t week2_a week2_d week2_t
# 1  1       1      22      10      12       4     -13      17
# 2  1       2      32      17      15       5     -13      18
# 3  1       3      12     -12      24       6     -23      29
# 4  2       4      11     -34      45       8     -11      19
# 5  2       5      33      10      23      10     -10      20