reaggregate: Aggregate or disaggregate at matrix

In travelr: Functions for Travel Demand Modeling

Description

Change the size of a matrix by adding or subdividing its cells.

Usage

reaggregate.matrix(mtx, eq, nrow=NULL, ncol=NULL)

Arguments

mtx	A matrix that will be refactored
eq	A data.frame which is a correspondence table describing how to compute output cells from input cells
nrow	An explict number of rows in the result matrix; the default is to use the maximum value of eq$o
ncol	An explict number of columns in the result matrix; the default is to use the maximum value of eq$p

Details

reaggregate.matrix will compute a new matrix with revised dimensions from an input matrix, mtx, based on the correspondences described in the equivalence table, eq.

The equivalence table must be a data.frame or matrix containing the following columns:

i

The row index of an input matrix cell

j

The column index of an input matrix cell

o

The row index of an output matrix cell

p

The column index of an output matrix cell

fact

The fraction of cell mtx[i, j] to be placed in cell mtx.out[o, p]

Each row in the equivalence table describes how to move some or all of the value in the [i,j] cell in the input matrix to the corresponding $[o,p] cll in the output matrix. The value in the input cell is multiplied by eq$fact and all the values contributing to the value of cell $[o,p] in the output are added together to form the final result.

If a particular i,j pair is not represented in the equivalence table then that pair will not contribute to the output matrix. Indices that are beyond the dimensions of the input matrix may introduce NA's into an output cell, even if other correspondences introduce known values.

With suitable correspondence factors, matrices may be expanded or reduced in size.

Value

A matrix with dimensions [nrow,ncol] computed by applying correspondence factors.

Author(s)

Jeremy Raw

Examples

   # The basic mechanics of aggregating a matrix...
   m <- matrix(1:100,nrow=10,ncol=10)
   eq <- expand.grid(i=1:10,j=1:10)
   eq$fact <- 1
   eq$o <- ceiling(eq$i/2)
   eq$p <- ceiling(eq$j/2)
   om <- reaggregate.matrix(m,eq)

   # Disaggregating is trickier since factors are necessary
   req <- eq[,c("i","j","fact","o","p")]
   names(req) <- c("o","p","fact","i","j")
   req$fact <- 0.25
   rom <- reaggregate.matrix(om,req)

   # Building an equivalence table using a correspondence table
   # Used in travel modeling to convert zones to districts
   ct <- data.frame( from=c( 1, 2, 3, 4, 5, 6, 7, 8, 9,10),
                       to=c( 2, 2, 1, 3, 3, 3, 4, 4, 5, 5) )
   c.eq <- expand.grid(i=1:10,j=1:10)
   c.eq$o <- ct$to[match(c.eq$i,ct$from)]
   c.eq$p <- ct$to[match(c.eq$j,ct$from)]
   c.eq$fact <- 1
   c.om <- reaggregate.matrix(m,c.eq)

travelr documentation built on May 2, 2019, 5:17 p.m.