specmine (version 4.0.0)

flat_pattern_filter: Flat pattern filter

Description

Performs a flat pattern filter over the dataset.

Usage

flat_pattern_filter(
  dataset,
  filter.function = "iqr",
  by.percent = TRUE,
  by.threshold = FALSE,
  red.value = 0
)

Value

A dataset object with the same overall structure as the input, where `dataset$data` has been filtered according to the selected flat-pattern criterion. Variables failing the selection rule are removed, the remaining dataset components are preserved, and `dataset$description` is updated to record the filtering step.

Arguments

dataset

A dataset object to filter.

filter.function

Filtering function to use. One of `"iqr"`, `"rsd"`, `"rnsd"`, `"sd"`, `"mad"`, `"mean"`, or `"median"`.

by.percent

Logical. If `TRUE`, the number of variables to filter will be defined as a percentage of the number of variables in the dataset; percentage is given by `red.value`.

by.threshold

Logical. If `TRUE`, filtering selects variables where the filtering function is above or equal to a threshold.

red.value

Reduction value. If `by.percent = TRUE`, this is the percentage of variables to remove, or `"auto"` for automatic calculation. If `by.threshold = TRUE`, this is the minimum value needed to keep the variable.

Examples

if (requireNamespace("specmine.datasets", quietly = TRUE)) {
  data(propolis, package = "specmine.datasets")
  propolis_proc = missingvalues_imputation(propolis)
  propolis_proc = flat_pattern_filter(propolis_proc, "iqr",
    by.percent = TRUE, red.value = 75)
}