checkts: Check and trim time-series

Description

Checks the length of two input time-series, and evaluates whether they have to be discarded given a specified threshold. If the difference between series is smaller than the threshold, the longest series between the two is either padded, or trimmed, to be the same length.

Usage

checkts(ts1, ts2, datatype, thrshd, pad)

Arguments

ts1

First time-series

ts2

Second time-series

datatype

Datatype of the time-series, either "numerical" or "categorical"

thrshd

Maximal length difference between the two-series used to accept them as valid. Series with a difference in length bigger than threshold are rejected as invalid.

pad

A logical flag indicating whether, in case the two-series differ by length, have to be padded (TRUE), with the shorter being extended. Or not, in which case, the longest series is trimmed to be of the same length as the shortest.

Value

If the difference between the two-series is smaller than the threshold, it returns a list with two arguments: the first [[1]] is a matrix with two columns, i.e., the two time-series padded or trimmed, and number of rows equal to the shortest sequence. The second [[2]] is a boolean flag with value TRUE, indicating that the difference was shorter than the threshold. If instead the difference was bigger than threshold, it returns a list with two arguments: the first [[1]] is the difference observed between the two series, the second [[2]] is a boolean flag with value FALSE, indicating that the difference was bigger than threshold.

Details

This function strictly applies when the two time-series series are expected to have equal length. An example of such case is co-registred eye-movement responses, where a difference in length highlights a possible error in the eye-tracking co-registration. The threshold is used to discriminate small versus large differences. The value to assign to the threshold should be estimated by looking at the difference distribution observed in the dataset. Note, when the two series are of continous numerical values, the padding value is obtained as the mean between the two series.

Examples

Run this code

# NOT RUN {
## generate two time-series of different length

ts1 = seq(1,30,1); ts2 = seq(1,25,1)
datatype = "continuous"
threshold = 6 ## threshold is larger than difference
pad = FALSE

res = checkts(ts1, ts2, datatype, threshold, pad)
print(res)

threshold = 4 ## threshold is smaller than difference

res = checkts(ts1, ts2, datatype, threshold)
print(res)

# }