align: Adjust ERP curves by time warping

In tdeenes/eegR: Analyze EEG Signals

Description

align transforms the ERP curves by aligning them on the time dimension.

Usage

align(erp, w, time_dim = "time")

Arguments

erp	the numeric vector or array of ERPs
w	the numeric vector or array of the warping weights (see warp)
time_dim	either a numeric or a character index indicating the time dimension of the ERP array (default: "time")

Details

This function is useful if the warping weights are computed on a specific subset of the ERP data, but the whole array should be aligned using these pre-computed weights. So the standard analysis pipeline is to call warp on the subset of the ERPs, store the computed weights in an array, and call align on the original (non-subsetted) array of ERPs and the stored warping weights.
In the simplest case, erp and w are vectors. They must be of equal length. In the second case, erp and w are arrays of equal dimensions. In the third case, erp is an array with more dimensions than w. In this scenario erp and w must have named dimensions, and all dimension identifiers in w must be present in erp.

See Also

warp

Examples

# load example data
data(erps)

# collapse stimclass and pairtype dimensions + downsample
x <- avgDims(erps, c("stimclass", "pairtype"))
x <- avgBin(x, "time", bin_length = 3L)
str(x) 

# let's assume we want to align the individual ERPs so that the individual
# variation in the peak latencies can be controlled for;
# we will use the Global Field Power instead of warping the ERPs for each
# channel separately
w <- warp(compGfp(x), align_dim = "id", verbose = TRUE) # takes some time

# now perform the alignment on the original data
x_aligned <- align(x, w)

# bind the original and the aligned arrays
x_all <- bindArrays(raw = x, aligned = x_aligned, along_name = "warp")

# plot the GFP of the raw and aligned curves
plotERParray(compGfp(x_all), sepdim = "id", 
             minus_up = FALSE, grid = c(2, 1))

# repeat the above analysis with lambda = 1 to demonstrate how 
# the lambda parameter affects the results
w <- warp(compGfp(x), align_dim = "id", lambda = 1) # takes some time

# now perform the alignment on the original data
x_aligned <- align(x, w)

# bind the new results to the previous array
x_all <- bindArrays(x_all, aligned_lambda1 = x_aligned, along_name = "warp")

# plot the GFP of the raw, the default aligned, and the new aligned curves
plotERParray(compGfp(x_all), sepdim = "id", 
             minus_up = FALSE, grid = c(3, 1))
             

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