How to dummyfy wiht one row per ID and multple dummies == 1? - r

Question

I've been working in something related to the dummyfying of a variable, but I have not been able to get the result I'm looking for. I'm sure there must be an easy solution for it, but was not able to find it.
My data set that looks like this:
TM_ID
APPLICATION_YEAR
EXPIRATION
DURATION
NICE
5
1870
1902
32
CLASE 34
6
1870
1891
21
CLASE 32
19
1902
1943
41
CLASE 34
19
1902
1943
41
CLASE 16
20
1876
1881
5
CLASE 34
21
1876
1877
12
CLASE 34
70
1877
1902
25
CLASE 16
70
1877
1902
25
CLASE 34
and I would like to dummyfy the column NICE to obtain one row per TM_ID and ones in the columns representing each of the levels in column NICE.
I tried with dummyVars(); it created one dummy for each level but there were still more than one observation per TM_ID. I tried as well with dcast() or pivot_wider(), but I have not been able to achieve what I need. It should look like this:
TM_ID
APPLICATION_YEAR
EXPIRATION
DURATION
CLASE 16
CLASE 32
CLASE 34
5
1870
1902
32
0
0
1
6
1870
1891
21
0
1
0
19
1902
1943
41
1
0
1
20
1876
1881
5
0
0
1
21
1876
1877
12
0
0
1
70
1877
1902
25
1
0
1
Were there is one only observation per ID and 1s en each of the corresponding columns of NICE levels.
The only way I found to do this is one-hot encoding first with dummyVars(), and then group_by() and mutate(), like:
group_by(TM_ID) %>%
mutate(NICE_1 = sum(NICE.1), NICE_10 = sum(NICE.10), NICE_11 = sum(NICE.11)......
But the problem with this solution is that I had to type each argument in the mutate, one per dummy (level of the original variable). What if there were hundreds of levels?
Thanks for your help.

Answer 1

You may use pivot_wider to get the data in wide format and use values_fn = length to dummify the NICE column.
library(dplyr)
library(tidyr)
res <- df %>%
arrange(NICE) %>%
pivot_wider(names_from = NICE, values_from = NICE,
values_fn = length, values_fill = 0)
res
# TM_ID APPLICATION_YEAR EXPIRATION DURATION `CLASE 16` `CLASE 32` `CLASE 34`
# <int> <int> <int> <int> <int> <int> <int>
#1 19 1902 1943 41 1 0 1
#2 70 1877 1902 25 1 0 1
#3 6 1870 1891 21 0 1 0
#4 5 1870 1902 32 0 0 1
#5 20 1876 1881 5 0 0 1
#6 21 1876 1877 12 0 0 1
data
df <- structure(list(TM_ID = c(5L, 6L, 19L, 19L, 20L, 21L, 70L, 70L
), APPLICATION_YEAR = c(1870L, 1870L, 1902L, 1902L, 1876L, 1876L,
1877L, 1877L), EXPIRATION = c(1902L, 1891L, 1943L, 1943L, 1881L,
1877L, 1902L, 1902L), DURATION = c(32L, 21L, 41L, 41L, 5L, 12L,
25L, 25L), NICE = c("CLASE 34", "CLASE 32", "CLASE 34", "CLASE 16",
"CLASE 34", "CLASE 34", "CLASE 16", "CLASE 34")), row.names = c(NA,
-8L), class = "data.frame")

Answer 2

using data.table
library(data.table)
dcast(setDT(df)[order(NICE)], ... ~ NICE, value.var = 'NICE', length)
TM_ID APPLICATION_YEAR EXPIRATION DURATION CLASE 16 CLASE 32 CLASE 34
1: 5 1870 1902 32 0 0 1
2: 6 1870 1891 21 0 1 0
3: 19 1902 1943 41 1 0 1
4: 20 1876 1881 5 0 0 1
5: 21 1876 1877 12 0 0 1
6: 70 1877 1902 25 1 0 1
data
df <- structure(list(TM_ID = c(5L, 6L, 19L, 19L, 20L, 21L, 70L, 70L
), APPLICATION_YEAR = c(1870L, 1870L, 1902L, 1902L, 1876L, 1876L,
1877L, 1877L), EXPIRATION = c(1902L, 1891L, 1943L, 1943L, 1881L,
1877L, 1902L, 1902L), DURATION = c(32L, 21L, 41L, 41L, 5L, 12L,
25L, 25L), NICE = c("CLASE 34", "CLASE 32", "CLASE 34", "CLASE 16",
"CLASE 34", "CLASE 34", "CLASE 16", "CLASE 34")), row.names = c(NA,
-8L), class = "data.frame")

Answer 3

Great, that's been of much help.
What I wonder is, what if we want to do the same for several variables. Actually, my data has this structure:
TM_ID APPLICATION_YEAR EXPIRATION DURATION NICE INDUSTRY
1: 5 1870 1902 32 CLASE 34 food
2: 6 1870 1891 21 CLASE 32 tobacco
3: 19 1902 1943 41 CLASE 34 tobacco
4: 19 1902 1943 41 CLASE 16 paper
5: 20 1876 1881 5 CLASE 34 chemical
6: 21 1876 1877 12 CLASE 34 arms
7: 70 1877 1902 25 CLASE 16 chemical
8: 70 1877 1902 25 CLASE 34 machinery
And I'd need:
TM_ID APPLICATION_YEAR EXPIRATION DURATION CLASE 16 CLASE 32 CLASE 34 food tobacco paper chemical arms machinery
5 1870 1902 32 0 0 1 1 0 0 0 0 0
6 1870 1891 21 0 1 0 0 1 0 0 0 0
19 1902 1943 41 1 0 1 0 1 1 0 0 0
20 1876 1881 5 0 0 1 0 0 0 1 0 0
21 1876 1877 12 0 0 1 0 0 0 0 1 0
70 1877 1902 25 1 0 1 1 0 0 1 0 1
I tried with different combinations in dcast(), but what I get is a column for each combination of NICE and INDUSTRY. Concretely, I tried to follow the example in the data.table vignettes:
## new 'cast' functionality - multiple value.vars
DT.c2 = dcast(DT.m2, family_id + age_mother ~ variable, value.var = c("dob", "gender"))
DT.c2
# family_id age_mother dob_1 dob_2 dob_3 gender_1 gender_2 gender_3
# 1: 1 30 1998-11-26 2000-01-29 <NA> 1 2 NA
# 2: 2 27 1996-06-22 <NA> <NA> 2 NA NA
# 3: 3 26 2002-07-11 2004-04-05 2007-09-02 2 2 1
# 4: 4 32 2004-10-10 2009-08-27 2012-07-21 1 1 1
# 5: 5 29 2000-12-05 2005-02-28 <NA> 2 1 NA
But it couldn't make it work for my case. I can imagine a work-around doing it separately and then cbind, but I'm sure there is a cleaner way.
Thanks again for your help.