sample_partitions: Randomly sample a partition matrix for partitioned GLS

In morrowcj/remotePARTS: Spatiotemporal Autoregression Analyses for Large Data Sets

Randomly sample a partition matrix for partitioned GLS

Description

Create a matrix whose columns contain indices of non-overlapping random samples.

Usage

sample_partitions(
  npix,
  npart = 10,
  partsize = NA,
  pixels = NA,
  verbose = FALSE
)

Arguments

npix	number of pixels in full dataset
npart	number of partitions to create
partsize	size of each partition
pixels	vector of pixel indexes to sample from
verbose	logical: TRUE prints additional info

Details

If both npart and partsize is specified, a partition matrix with these dimensions is returned. If only npart, is specified, partsize is selected as the largest integer possible that creates equal sized partitions. Similarly, if only npart = NA, then npart is selected to obtain as many partitions as possible.

Value

sample_partitions returns a matrix with partsize rows and npart columns. Columns contain random, non-overlapping samples from 1:npix

See Also

Other partitionedGLS: MC_GLSpart(), crosspart_GLS()

Examples

# dummy data with 100 pixels and 20 time points
dat.M <- matrix(rnorm(100*20), ncol = 20)

# 4 partitions (exhaustive)
sample_partitions(npix = nrow(dat.M), npart = 4)

# partitions with 10 pixels each (exhaustive)
sample_partitions(npix = nrow(dat.M), partsize = 10)

# 4 partitions each with 10 pixels (non-exhaustive, produces warning)
sample_partitions(npix = nrow(dat.M), npart = 4, partsize = 10)

# index of 50 pixels to use as subset
sub.indx <- c(1:10, 21:25, 30:62, 70:71)

# 5 partitions (exhaustive) from only the specified pixel subset
sample_partitions(npix = nrow(dat.M), npart = 5, pixels = sub.indx)

morrowcj/remotePARTS documentation built on Sept. 17, 2023, 5:42 p.m.