Skip to content

R package release 0.9.4

Compare
Choose a tag to compare
@scanisius scanisius released this 28 Jul 10:28
· 1 commit to master since this release

Added:

  • The new fdr.method argument of the pairwise.discover.test function selects the false discovery rate estimation method used for multiple testing correction. Its default value selects a Benjamini-Hochberg procedure adapted for discrete test statistics. As an alternative, the standard Benjamini-Hochberg procedure can be selected. While the latter is much faster, it is also more conservative than the discrete version. The discrete method was always used in previous versions and is still the recommended choice.

Changed:

  • Speed improvements in discover.matrix and pairwise.discover.test. discover.matrix finishes instantaneously for small to medium-sized data sets and takes only a few seconds for very large data sets. pairwise.discover.test (with discrete Benjamini-Hochberg) is about twice as fast as in previous versions.
  • The matrixStats package is no longer a dependency.

Fixed:

  • Re-enabled OpenMP, which distributes the computations in pairwise.discover.test across multiple CPUs. OpenMP was disabled since R 4.0.
  • Fixed the underlying Fortran code to make it compile with GNU Fortran >= 10. Due to these changes the minimum supported GNU Fortran version is now version 5.