Evaluation.KL: Calculates KL divergence between empirical pairwise...

In StatRank: Statistical Rank Aggregation: Inference, Evaluation, and Visualization

Description

Calculates KL divergence between empirical pairwise preferences and modeled pairwise preferences

Usage

Evaluation.KL(Data.pairs, m, Estimate, pairwise.prob = NA, prior = 0,
  nonparametric = FALSE, ...)

Arguments

Data.pairs	data broken up into pairs using Breaking function
m	number of alternatives
Estimate	estimation object from an Estimate function
pairwise.prob	Function that given two alternatives from the the Parameters argument, returns back a model probability that one is larger than the other
prior	prior weight to put in pairwise frequency matrix
nonparametric	indicator that model is nonparametric (default FALSE)
...	additional arguments passed to generateC.model

Value

the KL divergence between modeled and empirical pairwise preferences, thinking of the probabilities as a probability distribution over the (n choose 2) pairs

Examples

data(Data.Test)
Data.Test.pairs <- Breaking(Data.Test, "full")
m <- 5
Estimate <- Estimation.PL.GMM(Data.Test.pairs, m)
Evaluation.KL(Data.Test.pairs, m, Estimate, PL.Pairwise.Prob)

StatRank documentation built on May 1, 2019, 8:22 p.m.