Loop "ifelse" including NA

Question

I want to run a loop using ifelse() and including NA in R. Here is a example of my dataset named DATA:

[,1]   [,2]    [,3]    [,4]
ID1     NA      NA      NA
ID2     0       0       0
ID3     1       1       1
ID4     0       0       1
ID5     1       1       0
ID6     0       NA      1
ID7     NA      0       0
ID8     NA      1       1

Basically, I want to add a new column at the end using these criteria:

If there is at least one "1" in column 2, 3 or 4 => 1
If there is NO "1" in column 2, 3 or 4, but at least one "0" => 0
If there is only "NA" in column 2, 3 or 4 => NOTHING

I thought to use something like that, but it doesn't work. It returns:

"NA" if there is no "1", but at least one "0", instead of
"NOTHING"

Example for line 6:

*ifelse(any((DATA[6,c(2:4)])==1), "1", ifelse(any((DATA[6,c(2:4)])==0), "0",
"NOTHING"))*

I don't really want to return "NOTHING". In reality I want "NA", but to verify if a "NA" was the result of something that doesn't work, I used "NOTHING". I also want to loop the whole thing for each line in my data set, but I'm not very good at it either.

What should your example data look like afterward? also show that "new column" — Andre Elrico, Nov 16 '18 at 11:53
Please make your example reproducible: https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example — jogo, Nov 16 '18 at 11:54
If your problem got solved please [choose an answer](https://stackoverflow.com/help/someone-answers). — Andre Elrico, Nov 22 '18 at 11:12

Andre Elrico · Answer 1 · 2018-11-16T12:09:32.180

1

rowSums.na <-
    function(x, na.rm = T){

        suma <- rowSums(x, na.rm = na.rm)
        ind  <- apply(x, 1, function(el){ all(is.na(el)) })

        suma[ind] <- NA

        return(suma)
}

df1$newCol<- +(rowSums.na(df1[,2:4]) > 0)

#> df1$newCol
#[1] NA  0  1  1  1  1  0  1

df1 is your data.
normal rowSums/ sum returns 0 when all values are NA. A great mistake IMO. Eval:
1. rowSums(matrix(NA, 3, 3), na.rm = T)
2. rowSums.na(matrix(NA, 3, 3))
+( ... ) will convert the boolean into a numeric type. run +( c(TRUE, FALSE, TRUE) ) in R console.

edited Nov 16 '18 at 12:09

answered Nov 16 '18 at 11:57

Andre Elrico

10,956
6
50
69

I did not know about the use of `+()`. – RLave Nov 16 '18 at 12:00

Loop "ifelse" including NA

1 Answers1