wilcrofter/hitandrun.jl

A Julia package based on the paper, "Hit and Run as a Unifying Device." by P. Diaconis & H. Andersen, (2007), *Journal de la Société Française de Statistique*, 148(4):5-28.

Hit and Run

A Julia package based on the paper, "Hit and Run as a Unifying Device." by P. Diaconis & H. Andersen, (2007), Journal de la Société Française de Statistique, 148(4):5-28, [PDF] and its references.

This package is primarily a personal experiment, an attempt to gauge my own interest (and ability) in subjects which the first author, above, regards as revolutionary [PDF]. Since Julia is potentially revolutionary as well¹, the experiment seems apt.

I've chosen the Hit and Run paper mainly because it includes and references accessible examples such as generating random contingency tables, i.e., arrays with non-negative integer entries and fixed row and column sums².

The paper's promise of unification, "an attempt to make sense of the zoo of Monte Carlo techniques," also appeals. The appeal is not only the prospect of avoiding a steep learning curve, but because generality facilitates variation and experiment, hence (possibly) progress.

1. Julia: A Fresh Approach to Numerical Computing Jeff Bezanson, Alan Edelman, Stefan Karpinski, Viral B. Shah: "[Julia] allows programmers to write clear, high-level, generic and abstract code that closely resembles mathematical formulas ... yet produces fast, low-level machine code that has traditionally only been generated by static languages.”

2. I had originally expected to use models as specified in Mamba, an excellent, purely Julia platform for Monte Carlo Markov Chains. However, that specification, which is very natural for Bayesian networks, is somewhat unnatural for the examples to be considered here. Hence, I expect to use custom code to produce instances of the Chains type, to which Mamba's convergence diagnostics and other features apply.