This Matlab/Octave package implements EM and Gibbs sampling algorithms
for Bayesian inference in Bradley-Terry and Plackett-Luce
models (Caron
and Doucet, 2012).
These models are very popular to deal with data consisting of pairwise
or multiple comparisons of a set of items. Please refer to the
reference below for
further details on the models and algorithms.
The package has been tested on Matlab
R2014a (with statistics toolbox) and Octave 3.6.4.
F. Caron, A. Doucet. Efficient
Bayesian inference for generalized Bradley-Terry models. Journal of Computational and
Graphical Statistics, vol. 21(1), pp. 174-196,
2012.
[Download paper]
Installation
Download the package using the link above and unzip the
archive in some folder. Add the folder to the Matlab/Octave path using
the command addpath('path_to_the_folder');
Contents
Bayesian Inference in
Bradley-Terry models
test_bt.m:
test script for Bradley-Terry functions
btem.m:
EM algorithm for the Bradley-Terry model
btgibbs.m:
Gibbs sampler for the Bradley-Terry model
btemhome.m:
EM algorithm for the Bradley-Terry model with home advantage
btgibbshome.m:
Gibbs sampler for the Bradley-Terry model with home advantage
btemties.m:
EM algorithm for the Bradley-Terry model with ties
btgibbsties.m:
Gibbs sampler for the Bradley-Terry model with ties
btemhometies.m:
EM algorithm for the Bradley-Terry model with home advantage and ties
btgibbshometies.m:
Gibbs sampler for the Bradley-Terry model with home advantage and ties
Bayesian Inference in
Plackett-Luce models
test_plack.m:
test script for the Plackett-Luce functions
plackem.m:
EM algorithm for the Plackett-Luce model
plackgibbs.m:
Gibbs sampler for the Plackett-Luce model
Copyright
(C)
Copyright 2011 INRIA
Author: François Caron, Univ. of Oxford
caron@stats.ox.ac.uk
http://www.stats.ox.ac.uk/~caron/
Permission is granted for anyone to copy, use, or modify these programs
and accompanying documents for purposes of research or education,
provided this copyright notice is retained, and note is made of any
changes that have been made.
These programs and documents are distributed without any warranty,
express or implied. As the programs were written for research purposes
only, they have not
been tested to the degree that would be advisable in any important
application.
All use of these programs is entirely at the user's own risk.