Finding 3 columns with highest value in each row and put their name in a new data frame in R

Question

I have the following data frame:

temp1=data.frame(id=c(1,2,3,4,5),p1=c(0,0,1,1,2),p2=c(9,2,3,5,3),p3=c(1,4,10,3,7),p4=c(4,4,7,1,10))

  id p1 p2 p3 p4
  1  0  9  1  4
  2  0  2  3  4
  3  1  3 10  7
  4  1  5  3  1
  5  2  3  7 10

for each id I want to extract top three columns with highest value and put them in a new data frame something like this:

id top1 top2 top3
 1  p2   p4   p3
 2  p4   p3   p2
 3  p3   p4   p2
 4  p2   p3   p4/p1
 5  p4   p3   p2

in case there are two identical value we are allowed to put them in arbitrary order.

Answer 1

library("tidyr") 
library("dplyr")
df <- data.frame(id=c(1,2,3,4,5),p1=c(0,0,1,1,2),p2=c(9,2,3,5,3),p3=c(1,4,10,3,7),p4=c(4,4,7,1,10))
df2 <- gather(df,col,val,-id)   
res <- group_by(df2,id) %>% arrange(id,desc(val)) %>% summarise(top1 = first(col),top2 = nth(col,2),top3 = nth(col,3))

Results in

#      id  top1  top2  top3
#    <dbl> <chr> <chr> <chr>
#  1     1    p2    p4    p3
#  2     2    p3    p4    p2
#  3     3    p3    p4    p2
#  4     4    p2    p3    p1
#  5     5    p4    p3    p2

following new information

res <- group_by(df2,id) %>% mutate(r=rank(-(val/sum(val)*100),ties.method = "min")) %>% arrange(id,r) %>% summarise(top1 = first(col),top2 = nth(col,2),top3 = nth(col,3))

results in

#        id  top1  top2  top3
#      <dbl> <chr> <chr> <chr>
#    1     1    p2    p4    p3
#    2     2    p3    p4    p2
#    3     3    p3    p4    p2
#    4     4    p2    p3    p1
#    5     5    p4    p3    p2