Link data.frame and matrix - r

Question

I have a data.frame and a matrix with same row and different number of columns.
All elements in the matrix are integer but the data.frame includes character in some columns.
I want to link the rows of these file, i.e. if if I delete a row in the matrix the same row in the data.frame be deleted automatically or when I sort the elements of data.frame with one of its column, the elements in the matrix be sorted accordingly.
Added note: I want to keep the matrix as integer matrix so I can not use cbind.

Answer 1

There are (at least) two solutions to this. The easy option is to make a new data.frame which includes both rows as such:
Sample data
set.seed(123)
df <- data.frame(ID = 1:26, Group = sample(c("A", "B"), 26, TRUE))
mat <- matrix(rnorm(78), ncol = 3, dimnames = list(1:26, paste0("Val", 1:3)))
Make new data.frame, storing names of matrix columns for later reference:
new_df <- cbind(df, mat)
mat_cols <- colnames(mat)
Do some subsetting:
new_df <- new_df[seq(1, 25, 2), ]
Extract matrix back out whenever needed:
as.matrix(new_df[, mat_cols])
The other option is to use an S3 or S4 class. The Bioconductor package Biobase has, for example, an ExpressionSet class which can hold a matrix and phenotype data, and subsetting works to subset both (though the matrix has the rows and columns the opposite way round).
If you wanted to do that more simply (ExpressionsSets can be relatively complex to get your head around), here's an S3 implementation:
as.JoinedUp <- function(data_frame, matrix) {
stopifnot(is.data.frame(data_frame), is.matrix(matrix), nrow(data_frame) == nrow(matrix))
x <- list(data_frame = data_frame, matrix = matrix)
class(x) <- "JoinedUp"
x
}
`[.JoinedUp` <- function(x, i = NULL, j = NULL) {
if (is.null(i)) {
i <- 1:nrow(x$data_frame)
}
if (is.null(j)) {
j <- union(colnames(x$data_frame), colnames(x$matrix))
}
stopifnot(is.character(j))
x$data_frame <- x$data_frame[i, intersect(j, colnames(x$data_frame)), drop = FALSE]
x$matrix <- x$matrix[i, intersect(j, colnames(x$matrix)), drop = FALSE]
x
}
`[<-.JoinedUp` <- function(x, i = NULL, j = NULL, value) {
if (is.null(j)) {
j <- union(colnames(x$data_frame), colnames(x$matrix))
}
if (is.null(i)) {
i <- 1:nrow(x$data_frame)
}
stopifnot(is.character(j))
if (!is.matrix(value) & !is.data.frame(value)) {
value <- as.data.frame(t(value), stringsAsFactors = FALSE)
}
stopifnot(ncol(value) == length(j))
if (any(j %in% colnames(x$data_frame))) {
df_cols <- intersect(j, colnames(x$data_frame))
x$data_frame[i, df_cols] <- value[, match(df_cols, j)]
}
if (any(j %in% colnames(x$matrix))) {
mat_cols <- intersect(j, colnames(x$matrix))
x$matrix[i, mat_cols] <- data.matrix(value[, match(mat_cols, j)])
}
x
}
Examples:
new_obj <- as.JoinedUp(df, mat)
new_obj[1:3, ]
new_obj[, c("ID", "Val1")]
new_obj[10:15, ]$matrix
new_obj <- new_obj[order(new_obj$matrix[, "Val1"]), ]
new_obj[1:5, c("ID", "Val1")] <- data.frame(ID = 20:24, Val1 = 0)
This is only a skeleton of what you'd need; you'd probably also want to define methods for dim, nrow, ncol, etc.

Answer 2

Try this example:
#dummy data
set.seed(123)
df1 <- data.frame(ID=1:3, x=letters[1:3])
m1 <- matrix(c(1:3,runif(6)), ncol=3)
#cbind data.frame and matrix, results in a data.frame object
res <- cbind(df1, m1)
res
# ID x 1 2 3
# 1 1 a 1 0.2875775 0.8830174
# 2 2 b 2 0.7883051 0.9404673
# 3 3 c 3 0.4089769 0.0455565
#subset 2nd row
res[ 2,]
# ID x 1 2 3
# 2 2 b 2 0.7883051 0.9404673
#order by 4th column
res[ order(res[ ,4 ]), ]
# ID x 1 2 3
# 1 1 a 1 0.2875775 0.8830174
# 3 3 c 3 0.4089769 0.0455565
# 2 2 b 2 0.7883051 0.9404673