Post-calculation Principal Component Analysis

This function performs PCA on the input data frame to reduce number of predictors.

Usage

post_calc_pca(
  data,
  locs_id = "site_id",
  time_id = "time",
  yvar = "Arithmetic.Mean",
  coords = c("lon", "lat"),
  num_comp = 5,
  threshold = NA,
  pattern = "FUGITIVE|STACK",
  groups = NULL,
  prefix = "PCA",
  kernel = FALSE
)

Arguments

data

data.frame or data.table

locs_id

The column name in the spatial object that represents the location identifier.

time_id

The column name in the data frame that represents the time identifier.

yvar

The target variable.

coords

The column names that represent the XY coordinates. Default is c("lon", "lat").

num_comp

integer(1). The number of components to retain as new predictors. If threshold is defined, num_comp will be overridden.

threshold

numeric(1). A fraction of the total variance that should be covered by the components.

pattern

character(1). A regular expression pattern to match the columns that should be included in the PCA.

groups

character. A character vector of groups to perform PCA on. Each character should be a regular expression pattern to match the columns that should be included in the PCA. Default is NULL.

prefix

character(1). A prefix to be added to the column names of the Principal Components. Default is NULL.

kernel

logical(1). Whether to use a kernel PCA with recipes::step_kpca(). Default is FALSE.

Value

data.table with Principal Components sufficient to satisfy the threshold, merged with *_id and yvar columns from original data.

Note

If threshold is defined, num_comp will be overridden.

See also

reduce_pca() recipes::step_pca() recipes::step_kpca()