Create a new variable and assign the value to a group

Question

I have a tibble of 65524 observations where one variable is an ID for an household and the other is factor where the value of 1 is assigned if the age of the person in the household is less then 15y.o., 2 is assigned if the age is between 15 and 64, and 3 is assigned if the age of the person is 65 or older. The tibble look like this

> head(df, 15)
# A tibble: 15 x 2
   hh.id age.cat  
   <dbl> <dbl+lbl>
 1 11009 2        
 2 11009 2        
 3 11009 2        
 4 11009 2        
 5 11009 2        
 6 11009 1        
 7 11009 1        
 8 11009 1        
 9 11018 2        
10 11018 1        
11 11018 1        
12 11018 1        
13 11018 1        
14 11018 2        
15 11018 2

I need to create a variable to estimate the dependency ratio of each household. Something similar to this

 > head(df, 15)
# A tibble: 15 x 3
   hh.id age.cat  dep.ratio
   <dbl> <dbl+lbl><dbl>
 1 11009 2        0.60
 2 11009 2        0.60
 3 11009 2        0.60
 4 11009 2        0.60
 5 11009 2        0.60
 6 11009 1        0.60
 7 11009 1        0.60
 8 11009 1        0.60
 9 11018 2        1.25
10 11018 1        1.25
11 11018 1        1.25
12 11018 1        1.25
13 11018 1        1.25
14 11018 2        1.25
15 11018 2        1.25

I thought that using dplyr::mutate and dplyr::group_by would work

df <- df %>%
  dplyr::group_by(hh.id) %>%
  dplyr::mutate(dep.ratio = (length(which(df$age.cat == 1)) + length(which(df$age.cat == 3)))/length(which(df$age.cat == 2)))

However, I do not get the estimates per each group (ie per each household), but I get the overall dependency ratio for the whole sample, repeated for each observation.

# A tibble: 15 x 3
# Groups:   hh.id [2]
   hh.id age.cat   dep.ratio
   <dbl> <dbl+lbl>     <dbl>
 1 11009 2              1.02
 2 11009 2              1.02
 3 11009 2              1.02
 4 11009 2              1.02
 5 11009 2              1.02
 6 11009 1              1.02
 7 11009 1              1.02
 8 11009 1              1.02
 9 11018 2              1.02
10 11018 1              1.02
11 11018 1              1.02
12 11018 1              1.02
13 11018 1              1.02
14 11018 2              1.02
15 11018 2              1.02

I then considered using tapply , but I could not write a function which conditions on the values of hh.id . Finally, I also tried aggregate , but without any luck.

Any suggestion is welcome.

Thanks

Manolo

Answer 1

Here's an option:

ratiodf<- df %>% group_by(hh.id,age.cat) %>% 
  summarize(n=n()) %>% 
  spread(age.cat,n) %>% 
  mutate(ratio=(`1`+`3`)/`2`)

This will give you something like this:

# A tibble: 2 x 4
# Groups:   hh.id [2]
  hh.id   `1`   `2` ratio
  <int> <int> <int> <dbl>
1 11009     3     5  0.6 
2 11018     4     3  1.33

If you need to keep the data from the original df (eg, other columns), you can continue with a left_join:

left_join(df, ratiodf[,c(-2:-3)], by="hh.id")

Which will result in this:

   hh.id age.cat    ratio
1  11009       2 0.600000
2  11009       2 0.600000
3  11009       2 0.600000
4  11009       2 0.600000
5  11009       2 0.600000
6  11009       1 0.600000
7  11009       1 0.600000
8  11009       1 0.600000
9  11018       2 1.333333
10 11018       1 1.333333
11 11018       1 1.333333
12 11018       1 1.333333
13 11018       1 1.333333
14 11018       2 1.333333
15 11018       2 1.333333

The reason your code didn't work is that as soon as your put in the base format of df$hh.id etc., you bypass the tidyverse grouping and get the full column.