替换R中两个数据帧之间的值

Question

I am trying to solve a little question related to two dataframes in R. I have two data frames D1 and D2 :

D1
   ID I1 I2 I3
1 001  1  2  9
2 002  3  5  9
3 003  8  7  9
D2
   ID I1 I2 I3
1 001 NA  1 NA
2 002  1  1 NA
3 003 NA NA  1

These dataframes are produced with the next code:

D1=data.frame(ID=c("001","002","003"),I1=c(1,3,8),I2=c(2,5,7),I3=c(9,9,9),stringsAsFactors=FALSE)
D2=data.frame(ID=c("001","002","003"),I1=c(NA,1,NA),I2=c(1,1,NA),I3=c(NA,NA,1),stringsAsFactors=FALSE)

The dataframe D2 has rows with NA and non NA values. So I want to replace the non NA values in D2 with their respective value in D1 . For example in the first row of D2 the second column in non NA , so this value should be replaced with 2 from D2 . I tried building a matrix with the non NA values in D2 by using this code:

mm=!is.na(D2[-1]) 
      I1    I2    I3
[1,] FALSE  TRUE FALSE
[2,]  TRUE  TRUE FALSE
[3,] FALSE FALSE  TRUE

But when I trid to replace with a code like this D1[mm] I don't get the expected result. I would like to get something like this:

   ID I1 I2 I3
1 001 NA  2 NA
2 002  3  5 NA
3 003 NA NA  9

Thanks!

Answer 1

Try the following:

D2[!is.na(D2)] <- D1[!is.na(D2)]
D2
   ID   I1   I2   I3
1 001 <NA>    2 <NA>
2 002    3    5 <NA>
3 003 <NA> <NA>    9

Answer 2

@DatamineR's solution is the first one I thought of as well, but it has the unfortunate effect of indexing the data.frames as vectors, which promotes the atomic type to character (due to the ID column), which you may not want.

Here's an alternative that preserves the numericness of the I1-3 columns:

aggregate(.~ID,rbind(D1,D2),function(a) if (is.na(a[2])) NA else a[1],na.action=na.pass);
##    ID I1 I2 I3
## 1 001 NA  2 NA
## 2 002  3  5 NA
## 3 003 NA NA  9