using R I have a 10 rows X 6 columns matrix. I need to split it into submatrices throuhg gruoping couples of columns with no overlapping.
i.e. matrix has columns A,B,C,D,E,F and I need to extract 3 different matrices (or data.frames or whatever object within financial packages like zoo or timeSeries) formed by columns AB, CD and EF.
PS: matrix contains financial data series and any couple of columns has a date columns and a NAV column
Using some dummy data (note you must have a dataframe as otherwise R would not allow you to hold Date and numeric values in a matrix [unless they were all converted to characters or raw numeric representations])
set.seed(42)
df <- data.frame(A = Sys.Date() + 0:9, B = rnorm(10),
C = Sys.Date() - 0:9, D = rnorm(10),
E = Sys.Date() - 20:29, F = rnorm(10))
> head(df)
A B C D E F
1 2013-04-05 1.3709584 2013-04-05 1.3048697 2013-03-16 -0.3066386
2 2013-04-06 -0.5646982 2013-04-04 2.2866454 2013-03-15 -1.7813084
3 2013-04-07 0.3631284 2013-04-03 -1.3888607 2013-03-14 -0.1719174
4 2013-04-08 0.6328626 2013-04-02 -0.2787888 2013-03-13 1.2146747
5 2013-04-09 0.4042683 2013-04-01 -0.1333213 2013-03-12 1.8951935
6 2013-04-10 -0.1061245 2013-03-31 0.6359504 2013-03-11 -0.4304691
one easy way to do this is to form an index for the columns you want - here I chose the first column of each pair, 1, 3, 5, etc.
start <- seq(1, by = 2, length = ncol(df) / 2)
Then, we lapply over the indices in start and select from our data frame the ith and ith + 1 columns where i takes each index from start in turn (df[i:(i+1)])
sdf <- lapply(start, function(i, df) df[i:(i+1)], df = df)
which gives:
> sdf
[[1]]
A B
1 2013-04-05 1.37095845
2 2013-04-06 -0.56469817
3 2013-04-07 0.36312841
4 2013-04-08 0.63286260
5 2013-04-09 0.40426832
6 2013-04-10 -0.10612452
7 2013-04-11 1.51152200
8 2013-04-12 -0.09465904
9 2013-04-13 2.01842371
10 2013-04-14 -0.06271410
[[2]]
C D
1 2013-04-05 1.3048697
2 2013-04-04 2.2866454
....
> str(sdf)
List of 3
$ :'data.frame': 10 obs. of 2 variables:
..$ A: Date[1:10], format: "2013-04-05" "2013-04-06" ...
..$ B: num [1:10] 1.371 -0.565 0.363 0.633 0.404 ...
$ :'data.frame': 10 obs. of 2 variables:
..$ C: Date[1:10], format: "2013-04-05" "2013-04-04" ...
..$ D: num [1:10] 1.305 2.287 -1.389 -0.279 -0.133 ...
$ :'data.frame': 10 obs. of 2 variables:
..$ E: Date[1:10], format: "2013-03-16" "2013-03-15" ...
..$ F: num [1:10] -0.307 -1.781 -0.172 1.215 1.895 ...
An advantage of keeping the sub-data frames in a list is that you can apply a function or other operation to the sub-data frames using a loop or a tools like lapply or sapply for example.
Are you looking for something like this?
require("zoo")
##matrix with random entries
mat <- matrix(rnorm(60), nrow=10, ncol=6)
colnames(mat) <- LETTERS[1:6]
## optional: create zoo object
#mat <- as.zoo(mat)
##access columns
mat[,c("A", "B")]
mat[,c("C", "D")]
mat[,c("E", "F")]
Note that creating a zoo object for the mock data above is not necessary, but from your question it is a bit unclear how your data looks like.
