Viewed: 2 - Published at: a few seconds ago

[ Function for duplicate rows ]

I have a dataframe like below:

> df
     pat_id disease
[1,] "pat1" "dis1" 
[2,] "pat1" "dis1" 
[3,] "pat2" "dis0" 
[4,] "pat2" "dis5" 
[5,] "pat3" "dis2" 
[6,] "pat3" "dis2" 

How can I write a function to get a third variable which indicates for the same pat_id the disease variable is the same or not , like below?

> df
     pat_id disease var3
[1,] "pat1" "dis1"  "1" 
[2,] "pat1" "dis1"  "1" 
[3,] "pat2" "dis0"  "0" 
[4,] "pat2" "dis5"  "0" 
[5,] "pat3" "dis2"  "1" 
[6,] "pat3" "dis2"  "1" 

Answer 1


Try ave() for the groupings, and wrap the result from any(duplicated()), with as.integer(). Then bind with cbind(). Although I might recommend you use a data frame instead of a matrix here.

cbind(
    df, 
    var3 = ave(df[,2], df[,1], FUN = function(x) as.integer(any(duplicated(x)))
)
#      pat_id disease var3
# [1,] "pat1" "dis1"  "1" 
# [2,] "pat1" "dis1"  "1" 
# [3,] "pat2" "dis0"  "0" 
# [4,] "pat2" "dis5"  "0" 
# [5,] "pat3" "dis2"  "1" 
# [6,] "pat3" "dis2"  "1" 

For larger data, I would recommend converting to a data table. The syntax is actually a bit nicer too, and it will likely be faster.

library(data.table)
dt <- as.data.table(df)
dt[, var3 := if(any(duplicated(disease))) 1 else 0, by = pat_id]

which gives

   pat_id disease var3
1:   pat1    dis1    1
2:   pat1    dis1    1
3:   pat2    dis0    0
4:   pat2    dis5    0
5:   pat3    dis2    1
6:   pat3    dis2    1

where column classes will be more appropriate (char, char, int). Or you could use as.integer(any(duplicated(disease))) instead of if/else.

Answer 2


Slightly long-winded, but gives you a boolean third variable which is more easily tested. It also doesn't care about data types

> df <- data.frame(pat_id=c("pat1","pat1", "pat2", "pat2", "pat3", "pat3"), 
+                  disease=c("dis1","dis1","dis0","dis5","dis2","dis2"), 
+                  stringsAsFactors = F)
> counts<-apply(table(df), 1, function(x) sum(x!=0))
> df2<-data.frame(pat_id=names(counts), all_the_same=(counts==1))
> df3<-merge(df,df2)
> df3
  pat_id disease all_the_same
1   pat1    dis1         TRUE
2   pat1    dis1         TRUE
3   pat2    dis0        FALSE
4   pat2    dis5        FALSE
5   pat3    dis2         TRUE
6   pat3    dis2         TRUE
> sapply(df3, class)
      pat_id      disease all_the_same 
 "character"  "character"    "logical" 

This doesn't care how many of each combination you have and should leave your strings as strings - not factors.

Having the new column as a logical lets you more easily do queries such as finding all patients for which it is true

> unique(df3$pat_id[df3$all_the_same])
[1] "pat1" "pat3"

Answer 3


One option using dplyr

library(dplyr)
as.data.frame(df) %>%
     group_by(pat_id) %>%
     mutate(var3 = as.integer(n_distinct(disease)==1))
#  pat_id disease  var3
#   (chr)   (chr) (int)
#1   pat1    dis1     1
#2   pat1    dis1     1
#3   pat2    dis0     0
#4   pat2    dis5     0
#5   pat3    dis2     1
#6   pat3    dis2     1