Using group_by and mutate in dplyr package to create new factor variable by id variable

4
votes

I have a hierarchical data frame in long format, where each row represents relationships, of which many can belong to a single person. Here is code for a small example dataset: 

df <- data.frame(id = as.factor(c(1,1,1,2,2,3,4,4)),
             partner = c(1,2,3,1,2,1,1,2),
             kiss = as.factor(c("Yes", "No", "No", "No", "No", "Yes", "Yes", "No")))

  id partner kiss
1  1       1  Yes
2  1       2   No
3  1       3   No
4  2       1   No
5  2       2   No
6  3       1  Yes
7  4       1  Yes
8  4       2   No

I want to create a new factor variable in this dataset that indicates whether the person (indicated by the 'id variable) never kissed any of their 'partners'. In other words, if the person had a kiss with any of their partners the new variable would indicate 'Yes' — they never had a kiss with any partner. Here is what I think it should look like:

  id partner kiss neverkiss
1  1       1  Yes        No
2  1       2   No        No
3  1       3   No        No
4  2       1   No       Yes
5  2       2   No       Yes
6  3       1  Yes        No
7  4       1  Yes        No
8  4       2   No        No

Ideally, I would like to find a way to create this variable without reshaping the dataset. I also prefer to use the dplyr package. So far, I've thought about using the group_by, and mutate functions in this package to create this variable. However, i'm not sure what helper functions I can use to create my specific variable. I'm open to other ideas outside of the dplyr package, but that would be first prize for me.

2
rdplyr
df %>% group_by(id) %>% mutate(neverkiss = {if (any(kiss == "Yes")) "No" else "Yes"}) - shadow
You could consider using a logical column instead of Yes and No. Then, it would just be df %>% group_by(id) %>% mutate(neverkiss = all(kiss != "Yes")) - talat

2 Answers

8
votes

This should do it

require(dplyr)

df <- data.frame(id = as.factor(c(1,1,1,2,2,3,4,4)),
             partner = c(1,2,3,1,2,1,1,2),
             kiss = as.factor(c("Yes", "No", "No", "No", "No", "Yes", "Yes", "No")))

df_new <- df %>% 
   group_by(id) %>% 
   mutate("neverkiss" = {if (any(kiss == "Yes")) "No" else "Yes"})

df_new

If the new column should contain factors you have to ungroup first

df_new <- df %>% 
   group_by(id) %>% 
   mutate("neverkiss" = {if (any(kiss == "Yes")) "No" else "Yes"}) %>% 
   ungroup() %>% 
   mutate("neverkiss" = as.factor(neverkiss))

class(df_new$neverkiss)
[1] "factor"

The reason is that factors cant be combined:

a <- as.factor(c("Yes", "Yes", "Yes"))
b <- as.factor(c("No", "No", "No")) 

c(a, b) # meaningless

As grouping is still active mutate is basically building the vector neverkiss as a combination of vectors for each id (group) which results in a vector of just one level (in this case "No").

5
votes

We can also do it with data.table

library(data.table)
setDT(df)[, neverkiss := if(any(kiss=="Yes")) "No" else "Yes" , id]