mantel 2.2.2

• The veridical correlation is no longer included in the sample correlations.
• The sample standard deviation is now used in the case of stochastic tests.

mantel 2.2.1

Remove call to np.math, which is deprecated in Numpy > 1.25

mantel 2.2.0

• The test() function now returns a MantelResult object, providing greater flexibility in working with the results
• Added a plot() function for plotting the distribution of sample correlations against the veridical

mantel 2.1.0

New ignore_nans option, which makes it possible to handle missing values in one of the matrices

mantel 2.0.0

• Renamed the project to mantel and packaged it up properly for PyPI

Penelope Pitstop

v1.2 Penelope Pitstop

Added permutation enumeration. For small matrices, the program automatically enumerates all possible permutations to test against rather than testing against a random sample. This is faster for small matrices and gives a deterministic result.

v1.1 Road Runner

Now computes the empirical p-value for the upper or lower tail. Return values are now r, p, z.

v1.0 Speedy Gonzales

This release maximizes efficiency by minimizing the amount of computation done in the Monte Carlo loop. More specifically, rather than compute the entire correlation on every iteration, most of it can be computed outside the Monte Carlo loop (e.g. the residuals, sum-of-squares, etc.). In addition, the Y residuals will be the same set of n(n-1)/2 numbers on every iteration, just in a different permutation. This means you can simply permute the Y residuals on each iteration (in a way that only yields the permutations that could be generated from valid permutations of the distance matrix).