按列子集数据框并返回这些子集的列表

Question

I want to take a data frame like this one:

df <- data.frame(
  SortCol1 = rep(c("One", "Two", "Three", "Four"), times = 5),
  SortCol2 = rep(c("A", "B"), times = 10),
  Arb1 = rep(c(1,0,1,1,0), times = 4),
  Arb2 = rep(c(0,1,1,0,0), times = 4)
)

   SortCol1 SortCol2 Arb1 Arb2
1       One        A    1    0
2       Two        B    0    1
3     Three        A    1    1
4      Four        B    1    0
5       One        A    0    0
6       Two        B    1    0
7     Three        A    0    1
8      Four        B    1    1
9       One        A    1    0
10      Two        B    0    0
11    Three        A    1    0
12     Four        B    0    1
13      One        A    1    1
14      Two        B    1    0
15    Three        A    0    0
16     Four        B    1    0
17      One        A    0    1
18      Two        B    1    1
19    Three        A    1    0
20     Four        B    0    0

Then subset it by SortCol1 and SortCol2 to return a list of all subsetted data frames.

I have done something similar to this many times before using ddply when I want to apply a function to the Arb1 and Arb2 columns.

eg I know that

ddply(df, c("SortCol1", "SortCol2"), numcolwise(sum))

Will subset based on the two columns I want, and return a minimal frame which has those columns and the sum function applied.

What I want is rather than applying a function to those columns, just have each subset returned as an element of a list.

Pretend the function that does that is called ddply_list . I would hope for something akin to

ddply_list(df, c("SortCol1", "SortCol2"))

Which would return a list whose elements would be the data frames (which I have manually created for now):

df[df$SortCol1=="One" & df$SortCol2 == "A",]
   SortCol1 SortCol2 Arb1 Arb2
1       One        A    1    0
5       One        A    0    0
9       One        A    1    0
13      One        A    1    1
17      One        A    0    1

df[df$SortCol1=="Two" & df$SortCol2 == "B",]
   SortCol1 SortCol2 Arb1 Arb2
2       Two        B    0    1
6       Two        B    1    0
10      Two        B    0    0
14      Two        B    1    0
18      Two        B    1    1

etc for all combinations of SortCol1 and SortCol2 .

If there's a function list that already, perfect! If not, any advice for how to get towards this solution would be awesome!

The main bit I'm not sure on, is the simplest way to return all subsets of a data frame (subsetted by columns) as a list of data frames.

To put it in another way, the ddply documentation described the .fun argument as... function to apply to each piece . I think what I want is a way of just returning each 'piece' as an element of a list (preferably with the columns used for subsetting still attached).

Answer 1

Turns out it's very simple:

split(df, df[c("SortCol1", "SortCol2"], drop=TRUE)

Answer stolen from here: Automatically subset data frame by factor

Usage:

split(x, f, drop = FALSE, ...)

Where x is a vector or dataframe and y is a factor or list of factors for defining groups.