FDX 2.0.1

• Fix: kernel functions were not properly moved to internal namespace.
• Minor documentation corrections.
• Fixed bugs for discrete Lehmann-Romano procedure that could lead to incorrect results or to infinite loops.
• Changed maintainer e-mail address.

FDX 2.0.0

• New features:
  • discrete.GR(), discrete.LR(), discrete.PB() and their respective wrappers DGR(), DLR(), DPB(), NDGR() , NDLR() and NDPB() are now generic functions. The previously existing functionality is implemented in *.default methods.
  • These generic functions got *.DiscreteTestResults methods for processing DiscreteTestResults R6 class objects from package DiscreteTests directly, so they can be used within pipes. They also offer some performance gains because some time-consuming checks for consistency are no longer necessary.
  • For consistency of new generics and methods, the first parameter raw.pvalues needed to be renamed to test.results.
  • New parameter threshold for discrete.*(), continuous.*(), weighted.* and their respective wrapper functions. This enables selection of p-values which are smaller than or equal to a certain value. Note: observed p-values and their supports are then re-scaled, as the p-value distributions are now becoming conditional distributions. If no selection is performed (i.e. threshold = 1), print(), summary() and plot() outputs are as before. Otherwise, they now respect the re-scaled conditional distributions. Additionally, the FDX S3 class output objects of these functions now include a list Select with values and information regarding selection.
  • New parameter pCDFlist.indices for discrete.*() and their wrappers, which must have the same length as pCDFlist and may help increasing performance considerably. As pCDFlist may now include only unique supports, pCDFlist.indices must indicate the indices of the p-values which belong to a given support set. If pCDFlist has the same length as test.results, it can be omitted (by setting it to NULL, the default). If users prefer using DiscreteTestResults objects, they do not have to take care of this, as unique supports and indices are automatically extracted from these objects.
• New functions direct.discrete.GR(), direct.discrete.LR() and direct.discrete.PB() as more flexible replacements for fast.discrete.GR(), fast.discrete.LR() and fast.discrete.PB(). The latter have been marked as deprecated and will be removed in the future.
• Step function evaluation in C++ code has been replaced by closely optimized inline functions which offer performance gains of 10-50%.

FDX 1.0.6

• The “fix” of 1.0.5 could cause even more trouble and has been corrected.

FDX 1.0.5

• Fixed a severe bug that produced completely wrong critical constants of the continuous Guo-Romano and Lehmann-Romano procedures.

FDX 1.0.4

• Minor modification to kernel_DGR_fast C++ function to return approximated binomial probabilities, too.
• Added GitHub.

FDX 1.0.3

• Necessary updates due to changed signatures of imported functions of package PoissonBinomial.

FDX 1.0.2

• Minor improvements to internal functions to achieve better accuracy for upper tail Poisson binomial probabilities.

FDX 1.0.1

• Fixed a bug that prevented compilation on Solaris and clang configurations.

FDX 1.0.0

• Initial release.
• Added a NEWS.md file to track changes to the package.