Insert rows based on column entry and determine the entry of the inserted row based on where it inserts

Question

I have this data, which I named A:

A <- read.table(text = "ID  TIME    EVID    AMT DOSE
1   10      1       100 20
1   12      1       100 20
1   14      1       100 20
1   16      1       100 20
1   17      0       100 20
1   18      1       100 20
1   20      1       100 20
1   22      1       100 20
2   5       1       100 40
2   10      1       100 40
2   15      1       100 40
2   17      0       100 40
2   20      1       100 40
3   4       1       100 25
3   7       1       100 25
3   10      1       100 25
3   11      0       100 25
3   13      1       100 25
3   16      1       100 25
3   19      1       100 25", header = TRUE)

And my goal is insert new rows with EVID=2, ID the same as the preceding row ID, and TIME = preceding row's TIME entry plus AMT/DOSE, and I want the new rows to be follow after the first EVID=1 after the 0s, as below:

ID  TIME    EVID    AMT DOSE
1   10      1       100 20
1   12      1       100 20
1   14      1       100 20
1   16      1       100 20
1   17      0       100 20
1   18      1       100 20
1   23      2       100 20
1   20      1       100 20
1   22      1       100 20
2   5       1       100 40
2   10      1       100 40
2   15      1       100 40
2   17      0       100 40
2   20      1       100 40
2   22.5    2       100 40
3   4       1       100 25
3   7       1       100 25
3   10      1       100 25
3   11      0       100 25
3   13      1       100 25
3   17      2       100 25
3   16      1       100 25
3   19      1       100 25

I get as far as to indexing my EVID's

rle(as.character(EVID))$lengths
A$Index<-unlist(sapply(rle(as.character(EVID))$lengths, seq_len), use.names = FALSE)

In this circumstance this code works better than ave(EVID, EVID, FUN=seq_along) which would index all the 1s and all the 0s regardless of whether they are continuous. I want to insert my new rows between Index=1 and Index=2 rows (I will just manually delete the first new row).

   ID TIME EVID AMT DOSE Index
1   1   10    1 100   20     1
2   1   12    1 100   20     2
3   1   14    1 100   20     3
4   1   16    1 100   20     4
5   1   17    0 100   20     1
6   1   18    1 100   20     1
7   1   20    1 100   20     2
8   1   22    1 100   20     3
9   2    5    1 100   40     4
10  2   10    1 100   40     5
11  2   15    1 100   40     6
12  2   17    0 100   40     1
13  2   20    1 100   40     1
14  3    4    1 100   25     2
15  3    7    1 100   25     3
16  3   10    1 100   25     4
17  3   11    0 100   25     1
18  3   13    1 100   25     1
19  3   16    1 100   25     2
20  3   19    1 100   25     3

The resulting A has a new Index column; I want the new rows to be between Index 1 and 2, i.e. after row number 1, 6, 13, and 19 in this example.

I've come across solutions in which we can make a column vector, then insert the column as row into the data by defined row number. How do I add the rows based on column entry and determined some entries dynamically?

Thanks for your help!

it's not altogether clear how you determine where specifically to insert the new row. Can you elaborate please — Ricardo Saporta, Mar 07 '13 at 06:36

score 4 · Accepted Answer · answered Mar 07 '13 at 06:49

here is a solution with data.table It's really just two lines of code (with a bit of comments)

library(data.table)
ADT <- data.table(row=1:nrow(A), A, key="ID")

# just to give an idea of how we can Find the first 0 after the first 1, look at the output from this
ADT[, list(row, EVID,c(NA,diff(EVID)), c(NA,diff(EVID))==1)]

# identify afer which row to insert
# the values you want to change, assign using the `=`
# the values to keep, just call the variable name, no `=` sign
newRows <- ADT[c(NA,diff(EVID))==1, list(row=row+1, ID, TIME=TIME+AMT/DOSE, EVID=2, AMT, DOSE)]

# rbind the new rows with the original DT
# then reverse order by EVID, and order by row.  
# After ordering, remove the first column (`row`) since it is not needed
newA <- rbind(ADT, newRows)[order(EVID, decreasing=TRUE)][order(row)][, -1, with=FALSE]


### Results: 

 > newA
    ID TIME EVID AMT DOSE
 1:  1   10    1 100   20
 2:  1   12    1 100   20
 3:  1   14    1 100   20
 4:  1   16    1 100   20
 5:  1   17    0 100   20
 6:  1   18    1 100   20
 7:  1   23    2 100   20
 8:  1   20    1 100   20
 9:  1   22    1 100   20
10:  2    5    1 100   40
11:  2   10    1 100   40
12:  2   15    1 100   40
13:  2   17    0 100   40
14:  2   20    1 100   40
15:  2   22    2 100   40
16:  3    4    1 100   25
17:  3    7    1 100   25
18:  3   10    1 100   25
19:  3   11    0 100   25
20:  3   13    1 100   25
21:  3   17    2 100   25
22:  3   16    1 100   25
23:  3   19    1 100   25
    ID TIME EVID AMT DOSE

Insert rows based on column entry and determine the entry of the inserted row based on where it inserts

1 Answers1