v1.1.2

• Improved the sample_posterior multithreading so that the remainder is spread out among the worker threads instead of the entire workload being run on the head process.
• Fixed an issue where sampling from a tree with the n_ballots parameter set to a value fewer than the number of observed ballots replacement raises an error.

v1.1.1

Fixed a bug in the dirichlet_tree$sample_posterior method where specifying n_threads=x would result in x+1 threads spawning.

v1.1.0

This release adds a new parameter to the sample_posterior methods. This flag allows you to sample from the posterior distribution without reusing the ballots in the sample.

v1.0.3

This release fixes major bugs on systems using the clang compiler with libc++. We also improve documentation for the underlying Dirichlet and Multinomial distributions.

v1.0.0

This is the first major release of the package.

v0.4.1

• Added a few extra arguments surrounding Ballot IO, mostly for testing.
• Changed some internal behavior to add robustness and reduce the number of unnecessary warnings.
• Added lots of tests, increasing coverage to ~98%

v0.4

This release introduces an additional maxDepth parameter for the IRV tree structure. It also deprecates the marginal probability sampling functions due to instability and the fact it is unused in our analyses. In the future it will be replaced by a more extensive probabilistic toolkit for ballot data.

Dirichlet Tree auditing for IRV v0.3

This release includes all requirements for conducting Bayesian audits for IRV elections where ballots completely specify the candidate preference orderings.