ILSR: Estimate the parameters of a Bradley-Terry model
In Selbosh/scrooge: Statistical modelling of heterogeneous citation networks

Description Usage Arguments Details Value References See Also Examples

Estimate Bradley-Terry ability scores from paired comparison data using the I-LSR algorithm. Avoids using stats::glm() and seems to be faster, but only approximates maximum likelihood.

1	ILSR(C, sort = FALSE, maxits = 100, tolerance = 1e-06, verbose = FALSE)

`C`	a square matrix of paired comparisons
`sort`	logical. If `TRUE`, sort the weights in descending order
`maxits`	The maximum number of iterations in the I-LSR algorithm
`tolerance`	The criterion for convergence of the algorithm
`verbose`	logical. If `TRUE`, show the number of iterations as a message

The function uses iterative Luce Spectral Ranking (I-LSR) to approximate the maximum likelihood ability weights of the Bradley-Terry model. To do: Add unit tests.

A vector of estimated ability scores

Maystre, Lucas and Grossglauser, Matthias (2015). Fast and accurate inference of Plackett–Luce models. In Advances in Neural Information Processing Systems, 172–180.

Other network centrality estimators: BTscores, BradleyTerry, PageRank, Scroogefactor