apply: Apply Functions Over Sparse Matrix Margins

Description

Returns a vector or array or list of values obtained by applying a function to margins of a sparse matrix.

Usage

apply.spam(X, MARGIN=NULL, FUN, ...)

Arguments

the spam matrix to be used.

MARGIN

a vector giving the subscripts which the function will be applied over. 1 indicates rows, 2 indicates columns, NULL or c(1,2) indicates rows and columns.

FUN

the function to be applied.

...

optional arguments to FUN.

Value

Similar as a call to apply with a regular matrix. The most important cases are as follows. The result is a vector (MARGIN is length 1 and FUN is scalar) or a matrix (MARGIN is length 1 and FUN returns fixed length vectors, or MARGIN is length 2 and FUN is scalar) or a list (if FUN returns vectors of different lengths).

Details

This is a handy wrapper to apply a function to the (nonzero) elements of a sparse matrix. For example, it is possible to apply a covariance matrix to a distance matrix obtained by nearest.dist, see Examples.

A call to apply only coerces the sparse matrix to a regular one.

The basic principle is applying the function to @entries, or to the extracted columns or rows ([,i,drop=F] or [i,,drop=F]). It is important to note that an empty column contains at least one zero value and may lead to non intuitive results.

This function may evolve over the next few releases.

Examples

Run this code

# NOT RUN {
S <- as.spam(dist(1:5))
S <- apply.spam(S/2, NULL, exp)
# instead of 
# S@entries <- exp( S@entries/2) 

# Technical detail, a null matrix consists
# of one zero element.
apply.spam(S,c(1,2),pmax)
apply.spam(S,1,range)


# A similar example as for the base apply.
# However, no dimnames else we would get warnings. 
x <- as.spam(cbind(x1 = 3, x2 = c(0,0,0, 5:2)))
apply.spam(x, 2, mean, trim = .2)
col.sums <- apply.spam(x, 2, sum)
row.sums <- apply.spam(x, 1, sum)
rbind(cbind(x, row.sums), c(col.sums, sum(col.sums)))

apply.spam(x, 2, is.vector)

# Sort the columns of a matrix
# Notice that the result is a list due to the different
# lengths induced by the nonzero elements
apply.spam(x, 2, sort)

# Function with extra args:
cave <- function(x, c1, c2) c(mean(x[c1]), mean(x[c2]))
apply(x,1, cave,  c1=1, c2=c(1,2))

ma <- spam(c(1:4, 0, 0,0, 6), nrow = 2)
ma
apply.spam(ma, 1, table)  #--> a list of length 2
apply.spam(ma, 1, stats::quantile)# 5 x n matrix with rownames
# }

Run the code above in your browser using DataLab

Get 50% off unlimited learning