R：具有多个并行结果的分层ifelse条件

Question

I am looking for an elegant solution to the following problem:

I need to allocate owners to firms based on different matching criteria. These matching criteria are of different quality, hence criteria with weaker quality should only be used if higher quality criteria do not yield results. In my example, all a criteria are of the same quality level and of higher quality than b criteria.

The following to illustrate my point:

firmname <- c("Firm A", "Firm B", "Firm C", "Firm D", "Firm E", "Firm F")
ownermatch_a1 <- c("Owner 1", NA, NA, NA, "Owner 5", "Owner 6")
ownermatch_a2 <- c("Owner 1", NA, NA, "Owner 4", "Owner 5", "Owner 6")
ownermatch_a3 <- c("Owner 1", NA, "Owner 3", "Owner 4", "Owner 5", "Owner 6")
ownermatch_b1 <- c("Owner 1", "Owner 2", "Owner 3", "Owner 4", "Owner 5", "Owner 6")
ownerfinal <- (NA)

data.frame(firmname, ownermatch_a1, ownermatch_a2, ownermatch_a3, ownermatch_b1, ownerfinal)

This yields the following table

 firmname ownermatch_a1 ownermatch_a2 ownermatch_a3 ownermatch_b1 ownerfinal
1   Firm A       Owner 1       Owner 1       Owner 1       Owner 1       <NA>
2   Firm B          <NA>          <NA>          <NA>       Owner 2       <NA>
3   Firm C          <NA>          <NA>       Owner 3       Owner 3       <NA>
4   Firm D          <NA>       Owner 4       Owner 4       Owner 4       <NA>
5   Firm E       Owner 5       Owner 5       Owner 5       Owner 5       <NA>
6   Firm F       Owner 6       Owner 6       Owner 6       Owner 6       <NA>

I now want R to do the following: 1) If any Firm one of the 3 a criteria is non-NA, put it as ownerfinal . 2) If there are multiple parallel a non-NA, take any of these randomly and put as ownerfinal 3) Only if all of these are NA, take ownermatch_b1 and put as ownerfinal .

So in the example above: Firm A: Pick any of a1, a2, a3 Firm B: Pick b1 Firm C: Pick a3 Firm D: Pick either a2 or a3

Thanks!

Answer 1

No need for loops here. ?max.col is your friend for finding valid cases across columns and picking one at random:

tmp <- dat[2:4][cbind(seq_len(nrow(dat)), max.col(is.na(dat[2:4])))]
dat$ownerfinal <- replace(tmp, is.na(tmp), as.character(dat$ownermatch_b1)[is.na(tmp)])
dat

#  firmname ownermatch_a1 ownermatch_a2 ownermatch_a3 ownermatch_b1 ownerfinal
#1   Firm A       Owner 1       Owner 1       Owner 1       Owner 1    Owner 1
#2   Firm B          <NA>          <NA>          <NA>       Owner 2    Owner 2
#3   Firm C          <NA>          <NA>       Owner 3       Owner 3    Owner 3
#4   Firm D          <NA>       Owner 4       Owner 4       Owner 4    Owner 4
#5   Firm E       Owner 5       Owner 5       Owner 5       Owner 5    Owner 5
#6   Firm F       Owner 6       Owner 6       Owner 6       Owner 6    Owner 6

You could also use pmax if you want to take the first valid result:

do.call(pmax, c(lapply(dat[2:5],as.character), na.rm=TRUE) )
#[1] "Owner 1" "Owner 2" "Owner 3" "Owner 4" "Owner 5" "Owner 6"

Answer 2

doLookup <- function(x){
  for(i in 2:5){
    if(!is.na(x[i]))
      return(x[i])
  }
  return(NA)
}

#loop through each record and make assignment
for(j in 1:nrow(df))
   df[j,6] <- doLookup(df[j,])
df
  firmname ownermatch_a1 ownermatch_a2 ownermatch_a3 ownermatch_b1 ownerfinal
1   Firm A       Owner 1       Owner 1       Owner 1       Owner 1    Owner 1
2   Firm B          <NA>          <NA>          <NA>       Owner 2    Owner 2
3   Firm C          <NA>          <NA>       Owner 3       Owner 3    Owner 3
4   Firm D          <NA>       Owner 4       Owner 4       Owner 4    Owner 4
5   Firm E       Owner 5       Owner 5       Owner 5       Owner 5    Owner 5
6   Firm F       Owner 6       Owner 6       Owner 6       Owner 6    Owner 6