rowAvgsPerColSet: Applies a row-by-row (column-by-column) averaging function to...

In matrixStats: Functions that Apply to Rows and Columns of Matrices (and to Vectors)

Applies a row-by-row (column-by-column) averaging function to equally-sized subsets of matrix columns (rows)

Description

Applies a row-by-row (column-by-column) averaging function to equally-sized subsets of matrix columns (rows). Each subset is averaged independently of the others.

Usage

rowAvgsPerColSet(X, W = NULL, rows = NULL, S, FUN = rowMeans, ...,
  na.rm = NA, tFUN = FALSE)

colAvgsPerRowSet(X, W = NULL, cols = NULL, S, FUN = colMeans, ...,
  na.rm = NA, tFUN = FALSE)

Arguments

X	A numeric NxM matrix.
W	An optional numeric NxM matrix of weights.
rows	A vector indicating subset of rows to operate over. If NULL, no subsetting is done.
S	An integer KxJ matrix specifying the J subsets. Each column holds K column (row) indices for the corresponding subset.
FUN	The row-by-row (column-by-column) function used to average over each subset of X. This function must accept a numeric NxK (KxM) matrix and the logical argument na.rm, and return a numeric vector of length N (M).
...	Additional arguments passed to then FUN function.
na.rm	(logical) Argument passed to FUN() as na.rm = na.rm. If NA (default), then na.rm = TRUE is used if X or S holds missing values, otherwise na.rm = FALSE.
tFUN	If TRUE, the NxK (KxM) matrix passed to FUN() is transposed first.
cols	A vector indicating subset of columns to operate over. If NULL, no subsetting is done.

Details

If argument S is a single column vector with indices 1:N, then rowAvgsPerColSet(X, S = S, FUN = rowMeans) gives the same result as rowMeans(X). Analogously, for colAvgsPerRowSet().

Value

Returns a numeric JxN (MxJ) matrix, where row names equal rownames(X) (colnames(S)) and column names colnames(S) (colnames(X)).

Author(s)

Henrik Bengtsson

Examples

X <- matrix(rnorm(20 * 6), nrow = 20, ncol = 6)
rownames(X) <- LETTERS[1:nrow(X)]
colnames(X) <- letters[1:ncol(X)]
print(X)


# - - - - - - - - - - - - - - - - - - - - - - - - - -
# Apply rowMeans() for 3 sets of 2 columns
# - - - - - - - - - - - - - - - - - - - - - - - - - -
nbr_of_sets <- 3
S <- matrix(1:ncol(X), ncol = nbr_of_sets)
colnames(S) <- sprintf("s%d", 1:nbr_of_sets)
print(S)

Z <- rowAvgsPerColSet(X, S = S)
print(Z)

# Validation
Z0 <- cbind(s1 = rowMeans(X[, 1:2]),
            s2 = rowMeans(X[, 3:4]),
            s3 = rowMeans(X[, 5:6]))
stopifnot(identical(drop(Z), Z0))


# - - - - - - - - - - - - - - - - - - - - - - - - - -
# Apply colMeans() for 5 sets of 4 rows
# - - - - - - - - - - - - - - - - - - - - - - - - - -
nbr_of_sets <- 5
S <- matrix(1:nrow(X), ncol = nbr_of_sets)
colnames(S) <- sprintf("s%d", 1:nbr_of_sets)
print(S)

Z <- colAvgsPerRowSet(X, S = S)
print(Z)

# Validation
Z0 <- rbind(s1 = colMeans(X[  1:4, ]),
            s2 = colMeans(X[  5:8, ]),
            s3 = colMeans(X[ 9:12, ]),
            s4 = colMeans(X[13:16, ]),
            s5 = colMeans(X[17:20, ]))
stopifnot(identical(drop(Z), Z0))


# - - - - - - - - - - - - - - - - - - - - - - - - - -
# When there is only one "complete" set
# - - - - - - - - - - - - - - - - - - - - - - - - - -
nbr_of_sets <- 1
S <- matrix(1:ncol(X), ncol = nbr_of_sets)
colnames(S) <- sprintf("s%d", 1:nbr_of_sets)
print(S)

Z <- rowAvgsPerColSet(X, S = S, FUN = rowMeans)
print(Z)

Z0 <- rowMeans(X)
stopifnot(identical(drop(Z), Z0))


nbr_of_sets <- 1
S <- matrix(1:nrow(X), ncol = nbr_of_sets)
colnames(S) <- sprintf("s%d", 1:nbr_of_sets)
print(S)

Z <- colAvgsPerRowSet(X, S = S, FUN = colMeans)
print(Z)

Z0 <- colMeans(X)
stopifnot(identical(drop(Z), Z0))

matrixStats documentation built on Sept. 11, 2024, 5:24 p.m.