2

in a grouped data frame, i would like to apply a function that relates the one value in the actual a row to all other values of the group (and same column) except the one iin the present row. This will lead to a single value new variable. So if the group consists of c(1,2,3,4,5) i would like to have a new variable with: c(fun(1,c(2,3), fun(2, c(1,3), fun(3, c(1,2)) My groups do not have similar size. But trying so long, i always receive funny values like zeroes or errors.

Example code:

  set.seed(3)  
dat <- data_frame(a=1:10,value=round(runif(10),2),group=c(1,1,1,2,2,3,3,3,3,4))

 # one possible function
dif.dist <- function(x1, x2) sum(abs(x1 - x2))/(length(x2)-1) 

 # with this, sometimes the grouping gets lost in "vec" and i receive zeros   
 x <- dat%>%
 group_by(group)%>%
 mutate(vec= list(value))%>%
 mutate(dif = dif.dist(unique(value),unlist(vec)[unlist(vec)!=value]))%>%
 ungroup()

 # another try with plyr, that returns only 0   
 dat <- ddply(dat, .(group), mutate, dif=dif.dist1(value[a==a],value[value!=value[a==a]]))

but the function works

  dif.dist(dat$value[1],dat$value[2:3])
 [1] 0.85

Later, i need this to receive distance matrices of a large set of variables related to each participant. I would be thankful for Help!

Sans Soleil
  • 43
  • 4

1 Answers1

1

One option would be to loop over the sequence of rows after grouping by 'group' and subset the elements of 'value' based on the index

library(dplyr)
library(purrr)
out <- dat %>%
         group_by(group) %>% 
         mutate(dif = map_dbl(row_number(), ~ dif.dist(value[.x], value[-.x])))

head(out, 2)
# A tibble: 2 x 4
# Groups:   group [1]
#      a value group   dif
#  <int> <dbl> <dbl> <dbl>
#1     1  0.17     1  0.85
#2     2  0.81     1  1.07
akrun
  • 874,273
  • 37
  • 540
  • 662
  • Thank you very much, i did not know purrr. Although: when i try it, an error message shows: "Fehler in rank(x, ties.method = "first", na.last = "keep") : Argument "x" is missing (ohne Standardwert)". Did you add sth to the data? Or do i need to fill sth in for x? Thank you for a hint! – Sans Soleil Jun 24 '18 at 07:00
  • @SansSoleil I was using only your data. Didn't get any error. May be due to a difference in version of the package? – akrun Jun 24 '18 at 15:36
  • 1
    yes, it was necessary to update, now everything runs perfectly- Thank you again! – Sans Soleil Jun 24 '18 at 19:47