CODEWITHSUNDEEP
rtidyr
strangeloop
strangeloop

Reputation: 193

partially spread out tidy data by repeated measures

I have some data structured like this:

structure(list(subject = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L), group = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L), .Label = c("group1", "group2"), class = "factor"), measurement = c("color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time", "color", "time"), item_pos = c("1", "1", "2", "2", "3", "3", "4", "4", "1", "1", "2", "2", "3", "3", "4", "4", "1", "1", "2", "2", "3", "3", "4", "4", "1", "1", "2", "2", "3", "3", "4", "4"), value = c("blue", "1508", "orange", "752", "black", "585", "red", "842", "red", "879", "white", "1455", "green", "1757", "orange", "2241", "white", "2251", "yellow", "1740", "red", "1962", "yellow", "1854", "green", "1859", "blue", "2156", "yellow", "2494", "green", "1757"), item = c("A", "A", "B", "B", "B", "B", "A", "A", "A", "A", "B", "B", "B", "B", "A", "A", "C", "C", "C", "C", "D", "D", "D", "D", "C", "C", "C", "C", "D", "D", "D", "D")), .Names = c("subject", "group", "measurement", "item_pos", "value", "item"), row.names = c(NA, -32L), class = "data.frame")

Which has multiple observations by subject by item, so the data for subject 1 looks like this:

> filter(df.tidy, subject==1)
  subject  group measurement item_pos  value item
1       1 group1       color        1   blue    A
2       1 group1        time        1   1508    A
3       1 group1       color        2 orange    B
4       1 group1        time        2    752    B
5       1 group1       color        3  black    B
6       1 group1        time        3    585    B
7       1 group1       color        4    red    A
8       1 group1        time        4    842    A

So within a group each item appears twice, and for each occurrence there is a measurement of color and time. The order in which items appear is in item_pos.

While I like this long format, a colleague needs it slightly 'wider', with the repeated color and time measures in their own colums by item. The desired format would be as follows:

  subject  group item color1 color2 time1 time2
        1 group1    A   blue    red  1508   842
        1 group1    B orange  black   752   585
...
        4 group2    D yellow  green  2494  1757

My feeling is that this ought to be possible using a combination of gather(), spread() and other dplyr verbs, but I'm not sure what the dplyr equivalent here would be for (in for-loop speak) looping through the items by group and collecting the color and time observations in subsequent columns. Help much appreciated!

Related questions I consulted:

Upvotes: 2

Views: 204

Answers (1)

akrun
akrun

Reputation: 887501

We can try dcast from library(data.table). Convert the 'data.frame' to 'data.table' (setDT(df.tidy), grouped by 'subject', 'measurement' and 'item', create a sequence column "N" and then use dcast to convert from 'long' to 'wide' format.

library(data.table)
setDT(df.tidy)[, N:=1:.N, by = .(subject, measurement, item)]
dcast(df.tidy, subject+group + item ~measurement + N, value.var="value", sep="")
#   subject  group item color1 color2 time1 time2
#1:       1 group1    A   blue    red  1508   842
#2:       1 group1    B orange  black   752   585
#3:       2 group1    A    red orange   879  2241
#4:       2 group1    B  white  green  1455  1757
#5:       3 group2    C  white yellow  2251  1740
#6:       3 group2    D    red yellow  1962  1854
#7:       4 group2    C  green   blue  1859  2156
#8:       4 group2    D yellow  green  2494  1757

Or using dplyr/tidyr, we group by the same column, create a sequence column ("N"), ungroup, paste the 'measurement' and 'N' columns to create 'measurementN' (using unite) and then spread the data to 'wide' format.

library(dplyr)
library(tidyr)
df.tidy %>%
    group_by(subject, measurement, item) %>% 
    mutate(N = row_number()) %>%
    ungroup() %>% 
    unite(measurementN, measurement, N, sep='') %>%
    select(-item_pos) %>% 
    spread(measurementN, value)
#  subject  group  item color1 color2 time1 time2
#    (int) (fctr) (chr)  (chr)  (chr) (chr) (chr)
#1       1 group1     A   blue    red  1508   842
#2       1 group1     B orange  black   752   585
#3       2 group1     A    red orange   879  2241
#4       2 group1     B  white  green  1455  1757
#5       3 group2     C  white yellow  2251  1740
#6       3 group2     D    red yellow  1962  1854
#7       4 group2     C  green   blue  1859  2156
#8       4 group2     D yellow  green  2494  1757

Upvotes: 1

Related Questions