scaleArray: Scale array across arbitrary dimension(s)

In tdeenes/eegR: Analyze EEG Signals

Description

scaleArray centers and possibly scales the dimensions of an array across the dimension(s) chosen by the user

Usage

scaleArray(
  dat,
  by_dims = NULL,
  along_dims = NULL,
  center = TRUE,
  scale = TRUE,
  base_subset = NULL,
  center_subset = NULL,
  scale_subset = NULL,
  keep_dimorder = TRUE
)

Arguments

dat	numeric matrix or array
by_dims, along_dims	integer or character vector referring to the dimension(s) of dat across/along which centering/scaling should occur. See Details.
center	logical value if means should be subtracted (default: TRUE)
scale	logical value if scaling should occur (default: TRUE)
base_subset	argument passed to subsetArray before computing the mean and/or the scale
center_subset	the same as base_subset but only applies for the centering step
scale_subset	the same as base_subset but only applies for the scaling step
keep_dimorder	logical value; if TRUE (default), the order of the dimensions is kept intact, otherwise the channel dimension will be the first dimension in the resulting matrix or array

Details

scaleArray is theoretically similar to scale, and can be conceived of as if one merges the by_dims dimensions and also the along_dims dimensions of an array, and calls scale on the resulting matrix. Thus, the arguments by_dims and along_dims are redundant: use whatever is more convenient in the given use case (by_dims defines the rows, along_dims the columns).

Value

A numeric matrix or array of the same size as the input object

Examples

# example dataset
data(erps)

# center and scale the data to normalize the subjects (coded with id in the
# array), i.e. all subjects should have mean = 0 and SD = 1
scaled_erps <- scaleArray(erps, along_dims = "id")
means <- apply(scaled_erps, "id", mean)
sds <- apply(scaled_erps, "id", sd)
stopifnot(all.equal(unname(means), rep(0, length(means))))
stopifnot(all.equal(unname(sds), rep(1, length(means))))

# you might want to normalize each subject in each experimental condition
# based only on the pre-stimulus activity
prestim <- as.numeric(dimnames(erps)$time) < 0
base <- list(time = prestim)
scaled_erps <- scaleArray(erps, by_dims = c("chan", "time"),
                          base_subset = base)
means <- apply(subsetArray(scaled_erps, base),
               c("stimclass", "pairtype", "id"),
               mean)
sds <- apply(subsetArray(scaled_erps, base),
             c("stimclass", "pairtype", "id"),
             sd)
stopifnot(all.equal(as.vector(means), rep(0, length(means))))
stopifnot(all.equal(as.vector(sds), rep(1, length(sds))))

tdeenes/eegR documentation built on April 19, 2021, 4:17 p.m.