Replace NA with values in another row of same column for each group in r - values not unique within group

Question

I have a question very similar to a previous one but I am unable to generalize it to my case.

I have data that looks sort of like this

Within each ID, I have several Vis rows. The ones of interest to me are only a and b . The data is such that for each column in the data (V1...V7), if a is present, b is present and for all values of a , b is missing and vice versa. I would like to combine Vis's a and b for each ID group such that I have a single row (either a or b or even a new one, it doesn't really matter) without any missing data for any of the columns.

Answer 1

Based on the image showed, may be this helps. Here I am using actual NAs with only a couple of V columns.

We create a numeric index for column names that start with 'V' followed by numbers ('nm1'). Convert the 'data.frame' to 'data.table' ( setDT(df1) ), grouped by 'ID', we use Map , loop over the columns specified by the index 'nm1' ( SD[, nm1, with=FALSE] ) and the 'Vis' column, replace the 'V' column elements where the 'Vis' is either 'a' or 'b' by the non-NA element ( na.omit(x[.. ), and assign the output to the numeric index.

library(data.table)
nm1 <- grep('V\\d+',colnames(df1)) 

setDT(df1)[, (nm1):= Map(function(x,y) 
    replace(x, which(y %in% c('a', 'b')), na.omit(x[y %in% c('a', 'b')])), 
     .SD[,-1, with=FALSE], list(.SD[[1]])), ID]

We change the 'b' values to 'a'

 df1[Vis=='b', Vis := 'a']

and get the unique rows

 unique(df1)
 #   ID Vis V1 V2
 #1:  2   a  1  2
 #2:  2   c  4  5
 #3:  3   a  3  4
 #4:  4   a  2  3
 #5:  4   c  3  4
 #6:  4   d  1  1

data

df1 <- data.frame(ID= rep(c(2,3,4), c(3,2,4)), Vis=c('a', 'b', 'c', 'a', 
 'b', 'a', 'b', 'c', 'd'), V1= c(1, NA, 4, 3, NA, NA, 2, 3, 1), 
 V2= c(NA, 2, 5, 4, NA, 3, NA, 4, 1), stringsAsFactors=FALSE)

Answer 2

Just sum the values you need while removing NAs. There are more vectorized ways to do this, but the for loop is a bit clearer.

for(I in unique(df1$ID)) {
  df_sub <- subset(df1, df1$ID==I & df1$Vis %in% c("a", "b"))
  df1 <- subset(df1, df1$ID != I)
  new_row <- apply(df_sub[, -1:-2], 2, sum, na.rm=TRUE)
  df1 <- rbind(df1, c(ID=I, new_row))
}