NEWS

---

fasano.franceschini.test 2.2.3

• Minor documentation changes

fasano.franceschini.test 2.2.2 (2024-02-07)

• Minor documentation changes
• Added citation

fasano.franceschini.test 2.2.1 (2023-05-11)

• Fixed issue with p-value computation

fasano.franceschini.test 2.2.0 (2023-05-08)

• The confidence interval calculation was removed as it was not correct.
• The test statistic was replaced with a scaled version so it is now integer valued. This does not affect p-values.
• The test is now exact (i.e. the type I error rate is exactly alpha).

fasano.franceschini.test 2.1.1 (2022-11-11)

• Minor changes in documentation
• Slight update to the range tree vs. brute force cutoffs
• Fixed issue that prevented compilation on some machines

fasano.franceschini.test 2.1.0 (2022-09-29)

• The 'o' (optimize) method has been removed as the benchmarking itself proved to be prohibitively slow. Instead, the default behavior now is to infer which method is likely faster based on the sample sizes and dimension of the data.
• The cores argument has been removed. Use threads instead.
• For large dimensions, memory was becoming an issue with the brute force approach as a std::vector was used to tabulate the number of points in each octant, and the vector is of size 2^d. For larger dimensions, tabulation is now done using a std::unordered_map, the size of which scales with the sample sizes instead of dimension.
• A seeded version of the parallel permutation has been added. The seed argument is no longer ignored when threads is larger than 1.

fasano.franceschini.test 2.0.1 (2022-08-17)

• Minor speedups introduced, particularly for the brute force method.
• The default computation method is now 'o' for optimize: the test statistic is computed using both the range tree and brute force methods, and the faster of the two is used for the permutation test.

fasano.franceschini.test 2.0.0 (2022-05-29)

• Major overhaul of the package. Most computations are now done in C++ now rather than R.
• Testing has been extended from two dimensions to arbitrary dimensions.
• A second method for computing the test statistic has been added which has a loglinear time complexity rather than quadratic.
• The cores argument has been renamed threads. cores is still allowed for backwards compatibility but will be removed in the next release.
• A seed argument has been added to ensure reproducibility in the p-value estimation.
• A verbose argument has been added to print a progress bar during the permutation test.
• The p-value distributional approximation has been removed, as the approximation is only valid in two dimensions.
• A confidence interval for the p-value estimate is now returned.
• Points that are tied along certain axes no longer add fractional counts to the test statistic. See paper for exact details on test statistic computation.

fasano.franceschini.test 1.1.0 (2021-09-02)

• The nBootstrap argument is deprecated, and will be removed in the next version of this package. Use nPermute instead of nBootstrap to compute the empirical null distribution for your data.

fasano.franceschini.test 1.0.1 (2021-07-17)

• More efficient quadCount() calculation reducing runtime.
• Test result now of class Htest for improved integration with other packages.
• Bug fix in bootstrapping procedure ensuring p-value is bounded between 0 and 1.

fasano.franceschini.test 1.0.0 (2021-06-22)

The fasano.franceschini.test has been released into the wild!

• For a comprehensive analysis and discussion of the fasano.franceschini.test, see the accompanying paper.