Researchers have posted a new paper on arXiv that tackles a core problem in redistricting analysis: how to generate a representative ensemble of district plans. The work frames district plans as graph partitions and introduces a class of methods called Bonsai for sampling these partitions independently.
Independent sampling is central to the approach. Rather than relying on dependent draws from a Markov chain, the methods are designed to draw each plan separately from a reasonable probability distribution over the space of partitions. This allows for the construction of an ensemble of district plans that can be used for statistical comparison.
The abstract states that the authors compare the performance of their methods, though the specific results are not included in the available excerpt. As a preprint, the paper has not been peer-reviewed. The full details of the Bonsai methods and their evaluation remain to be examined.