emp_kl_div {representr}

Calculate the empirical KL divergence for a representative dataset as compared to the true dataset

Description

Calculate the empirical KL divergence for a representative dataset as compared to the true dataset

Usage

emp_kl_div(
  true_dat,
  rep_dat,
  categoric_vars,
  numeric_vars,
  l_m = 10,
  weights = rep(1, nrow(rep_dat))
)

Arguments

true_dat

The true dataset

rep_dat

A representative dataset

categoric_vars

A vector of column positions or column names for the categoric variables.

numeric_vars

A vector of column positions or column names for the numeric variables.

l_m

Approximate number of true data points to be in each bin for numeric variables. Default is 10.

weights

If weighted frequencies are desired, pass a vector weights of the same length as representative data points.

Details

This function computes the estimated the KL divergence of two samples of data using the empirical distribution functions for the representative data set and true data set with continuous variables transformed to categorical using a histogram approach with statistically equivalent data-dependent bins, as detailed in

Wang, Qing, Sanjeev R. Kulkarni, and Sergio Verdú. "Divergence estimation of continuous distributions based on data-dependent partitions." IEEE Transactions on Information Theory 51.9 (2005): 3064-3074.

Examples

data("rl_reg1")

## random prototyping
rep_dat_random <- represent(rl_reg1, identity.rl_reg1, "proto_random", id = FALSE, parallel = FALSE)

## empirical KL divergence
cat_vars <- c("sex")
num_vars <- c("income", "bp")
emp_kl_div(rl_reg1[unique(identity.rl_reg1), c(cat_vars, num_vars)],
           rep_dat_random[, c(cat_vars, num_vars)],
           cat_vars, num_vars)