Question

I have to select a bunch of data from a data frame depending on certain conditions. The data frame looks roughly like this:

  F1 F2 D1 D2
1 A1 B1  1  0
2 A1 B1  1  1
3 A1 B1  0  0
4 A1 B2  1  0
5 A1 B2  0  0
6 A2 B2  1  0
7 A2 B2  1  1

The Fx are factors, and the Dx are data values. What I have to do is the following:

1. Find rows with data values that match a specific pattern.
2. For each row that matches that pattern, find all rows that have the same factors
3. For each unique factor combination, apply some operation to all rows that have that combination

For example,

factors <- unique(data[D1==1 & D2 == 1, c("F1","F2")])

will give me step 1 and most of 2.

And with

data[data$F1 %in% factors$F1 & data$F2 %in% factors%F2,]

I’m getting closer to the solution, but with the example data above, this will select all rows. But rows 4 and 5 should not be selected, because they are not an exact match. How can I add in some sort of condition that required that the %in% matches happen on the same row?

I feel like this is something that should be a common operation and thus R probably has a clever way for doing this.

Any ideas???? Thanks.

Answer 1

You can use the indexing of the data.table package to select all rows that have to be manipulated.

data <- data.table(data,key="F1,F2")
data[unique(data[D1 == 1 & D2 == 1,list(F1,F2)])]