autoSelectVariables: Automatic selection of predictive variables for species...

In BlasBenito/SDMworkshop: Streamlining the Development of Species Distribution Models

Description

Applies biserialCorrelation, correlationDendrogram (with correlation threshold set to 0.5), and autoVIF to automatically select a set of non-correlated variables with the higher biserial correlation as possible.

Usage

autoSelectVariables(
  x,
  presence.column = "presence",
  variables = NULL,
  exclude.variables = NULL,
  plot = TRUE,
  label.size = 6
)

Arguments

x	A data frame with a presence column with 1 indicating presence and 0 indicating background, and columns with predictor values.
presence.column	Character, name of the presence column.
variables	Character vector, names of the columns representing predictors. If NULL, all numeric variables but presence.column are considered.
exclude.variables	Character vector, variables to exclude from the analysis.
plot	Boolean, if TRUE, prints last correlation dendrogram to test the final output.
label.size	Numeric, size of the dendrogram labels.

Value

A character vector with the names of the selected variables.

Author(s)

Blas Benito <[e-mail]>.

Examples

## Not run: 
data(virtualSpeciesPB)
selected.vars <- autoSelectVariables(
  x = virtualSpeciesPB,
  presence.column = "presence",
  exclude.variables = c("x", "y")
)
selected.vars
HH::vif(virtualSpeciesPB[, selected.vars])
cor(virtualSpeciesPB[, selected.vars])

## End(Not run)

BlasBenito/SDMworkshop documentation built on March 4, 2020, 4:16 a.m.