Replace NA in one row of data frame with values from other

Question

i want to replace NA in one row with values from another row, example data are:

group <-c('A','A_old')
year1<- c(NA,'20')
year2<- c(NA,'40')
year3<- c('20','230')
datac=data_frame(group,year1,year2,year3)

group <-c('A','A_old')
year1<- c('20','20')
year2<- c('40','40')
year3<- c('20','230')
finaldatac=data_frame(group,year1,year2,year3)

Original table is much larger so referring to each element one by one and assigning value is not possible..

Thanks!

For sake of argument below, i need to refer to the row values by their name as original table is big and i can not play around with only two rows. For example in table below, i would like to replace row 1 (group==A) with row 5 (group==E). Data are here:

group <-c('A','B','C','D','E','F','G')
year1<- c(NA,'100',NA,'200','300',NA,NA)
year2<- c(NA,'100',NA,'200','300','50','40')
year3<- c('20','100',10,'200','300','150','230')
data=data.frame(group,year1,year2,year3)

SO i want to get:

group <-c('A','B','C','D','E','F','G')
year1<- c('300','100',NA,'200','300',NA,NA)
year2<- c('300','100',NA,'200','300','50','40')
year3<- c('20','100',10,'200','300','150','230')
data=data.frame(group,year1,year2,year3)

Answer 1

A simple replacement in Base R can be done with:

data[1,is.na(data[1,])] <- data[5,is.na(data[1,])]

Taking in account that only need to know row indices of desired replacements.

Answer 2

Other than using fill or na.locf, you could do:

datac %>%
  group_by(grp = gsub("_.*", "", group)) %>%
  mutate_at(vars(contains("year")),
                 funs(.[!is.na(.)])) %>%
  ungroup() %>% select(-grp)

Output:

# A tibble: 2 x 4
  group year1 year2 year3
  <chr> <chr> <chr> <chr>
1 A     20    40    20   
2 A_old 20    40    230  

For your second example, you could do:

data %>%
  mutate_at(
    vars(contains("year")), 
    funs(
      case_when(
        group == "A" & is.na(.) ~ .[group == "E"],
        TRUE ~ .)
      )
    )

Output:

  group year1 year2 year3
1     A   300   300    20
2     B   100   100   100
3     C  <NA>  <NA>    10
4     D   200   200   200
5     E   300   300   300
6     F  <NA>    50   150
7     G  <NA>    40   230

You can also add other conditions to case_when.

For instance, if you'd additionally like to replace C years with what is there for group D, you would add:

data %>%
  mutate_at(
    vars(contains("year")), 
    funs(
      case_when(
        group == "A" & is.na(.) ~ .[group == "E"],
        group == "C" & is.na(.) ~ .[group == "D"],
        TRUE ~ .)
    )
  )

Answer 3

After a very long evening and headache from r i managed to get this:

rm(list = ls())
group <-c('A','A old')
year1<- c(NA,'20')
year2<- c(NA,'40')
year3<- c('20','230')
datac=data_frame(group,year1,year2,year3)
group <-c('A','A old')
year1<- c('20','20')
year2<- c('40','40')
year3<- c('20','230')
finaldatac=data_frame(group,year1,year2,year3)


datac$group <- gsub(' ', '--', datac$group)
datact = t(datac)
colnames(datact) = datact[1, ] 
datact = datact[-1, ] 

datact[,"A"]  <- ifelse(!is.na(datact[,"A"]), datact[,"A"] , datact[,"A--old"])
datactt=t(datact)
group = rownames(datactt)
datactt<-cbind(datactt, group)
rownames(datactt) <- c()
datactt <- as.data.frame(datactt)

sapply(datactt, class)
datactt <- data.frame(lapply(datactt, as.character), stringsAsFactors=FALSE)


datactt$group <- gsub('--', ' ', datactt$group)

Where datactt (hopefully) is the same as finaldatac that i wanted... I am sure this cant be the best solution, obviously not the prettiest. If anybody has something similar, but shorter or more efficient please post it i would appreciate the answer.