Select important variables based on stabilized correlations — select_important

This function selects important variables from a dataset based on stabilized correlations computed using the Sinkhorn method.

select_important_variables(
  data,
  cols,
  threshold = 0.5,
  epsilon = 0.001,
  max_iter = 0,
  use_glasso = 0.1,
  least = FALSE,
  return_raw_importance = FALSE
)

Arguments

data: A data frame containing the variables.
cols: A character vector specifying which columns (variables) to consider.
threshold: Proportion of variables to select based on importance (default: 0.2).
epsilon: Convergence threshold for Sinkhorn iterations (default: 1e-3).
max_iter: Maximum number of Sinkhorn iterations (default: 100).
use_glasso: set a scalar greater than zero to use graphical LASSO; this parameter relates to sparseness levels
least: boolean take least explanatory (most uncorrelated) variables
return_raw_importance: boolean

Value

A character vector of selected variable names.

Examples

set.seed(123)
data <- data.frame(
  x1 = rnorm(100),
  x2 = rnorm(100),
  x3 = rnorm(100),
  x4 = rnorm(100)
)
selected_vars <- select_important_variables(data, 
 c("x1", "x2", "x3", "x4"), threshold = 0.3)
print(selected_vars)
#> [1] "x1"