Skip to contents

KL Divergence Between Population and Sample Distributions

Source: R/kl_optimization.R
ss_kl_divergence.Rd

Computes the mean Kullback-Leibler divergence across all numeric variables shared by a population data set and a sample drawn from it.

Usage

ss_kl_divergence(population_data, sample_data, n_bins = 25)

Arguments

population_data

Data frame of population ancillary data.

sample_data

Data frame of sample ancillary data.

n_bins

Integer, number of histogram bins used to estimate each variable's distribution. Default 25.

Value

Numeric, the mean KL divergence across all numeric variables.

Details

For each numeric variable, population and sample distributions are estimated as histograms over the same bins (defined by the population range). KL divergence is then \(\sum O_i \log(O_i / E_i)\), where \(O_i\) is the sample density and \(E_i\) the population density in bin \(i\). The result is averaged across variables.

Examples

pop <- data.frame(x = rnorm(200), y = runif(200))
samp <- pop[sample(nrow(pop), 30), ]
ss_kl_divergence(pop, samp)
#> [1] 0.3444531