NEWS.md

In hsloot/rmo: A package for the Marshall-Olkin distribution

rmo 0.9

• Rename multiple S4-methods and arguments (#121):
• Rename S4-method exIntensities() to calcExShockSizeArrivalIntensities() and arguments/variables ex_intensities to theta
• Rename S4-method uexIntensities() to calcExShockArrivalIntensities()
• Rename S4-method exQMatrix() to calcMDCMGeneratorMatrix()
• Rename S4-method intensities() to calcShockArrivalIntensities()
• Rename S4-method defaultMethod() to getDefaultMethodString()
• Rename S4-method levyDensity() to getLevyDensity()
• Rename S4-method stieltjesDensity() to getStieltjesDensity()
• Rename S4-method valueOf() to calcIterativeDifference()
• Rename S4-method valueOf0() to calcValue()

• Rename various variable, method, and argument names appropriately

• Improve extensibility by exposing defaultMethod and valueOf0 (#118)

• Review and improve documentation (#116)

• Re-organize S4 code and documentation (#117)

• Remove outdated rmarkdown notebooks from development phase (#115)

• Remove convenience wrapper functions for generating shock arrival intensity and shock-size arrival intensity parameter (#114)

• Add ConvexCombinationOfBernsteinFunctions to represent convex combinations of Bernstein functions compactly

• Small changes of the documentation

• Improving implementation of method exIntensities

• Provide zero-sample-size parametrization, n = 0, in simulation algorithms (e.g. for measure setup time)

• Bugfix for consistent use of STL headers and attribute [[maybe_unused]] in C++ code

rmo 0.8

• Breaking change: Rename simulation method *_markovian to *_mdcm and *_arnold to *_am

• Breaking change: Provide a single entry point method for general MO sampling distributions rmo, exchangeable MO sampling routine rexmo, extendible MO sampling routines rextmo, and parametrized extendible MO sampling routines rpextmo

• Rename C++ backend classes arnold_mo_distribution, markovian_exmo_distribution, and armageddon_extmo_distribution to am_mo_distribution, mdcm_exmo_distribution, and esm_armextmo_distribution, respectively

• Improve code coverage by adding snapshot tests for sampling routines

• Improve documentation

rmo 0.7

• Breaking change: Remove lambda parameter for PoissonBernsteinFunction

rmo 0.6

• Breaking change: Rename concept Cuadras-Augé to armageddon ESM to better reflect the nature of the distribution. Associated functions are renamed accordingly

• Bugfix to handle integration error explicitly, possibly adjusting in case of for very small values

• Improve numerical stability by using the explicit first-order iterated difference for difference_order == 1L in valueOf to avoid endpoint-singularity problems in numerical integration

• Bugfix in initializers and validity methods of S4 objects such that validObject can now be called with the argument test = TRUE without causing an error

• Allow pass-though of arguments uexIntensities, exIntensities, and exQMatrix from method valueOf to integrate

• Add new CompositeScaledBernsteinFunction class

• Add show method for BernsteinFunction-classes.

rmo 0.5

• Breaking change: Rename function names of simulation algorithms. Now, sampling algorithms follow the format r*mo_* where the first * indicates the input-parameter and the second on the algorithm, e.g. rexmo_markovian has ex_intensities as input parameters and uses the Markovian model for the default counting process.

• Breaking change: Rearrange order in arguments of valueOf: instead of cscale, n, k we have n, k, cscale.

• Implementing new methods to generate distribution parameter from BernsteinFunction classes

rmo 0.4

• Change input parameter for rexmo_markovian (which is now is scaled exchangeable intensity). The ex_intensities*-methods are similarly adjusted such that there should be no change necessary if these functions were used to create the input parameter for rexmo_markovian.
• The Bernstein function classes have been refactored and some new features have been added.

rmo 0.3

• Breaking change: Rename rmo_ex_arnold to rexmo_markovian.

• Improve numerical stability of the calculation of products with large binomial coefficients

• Bugfix in function is_within to avoid undefined behavior

• Implementing the Inverse Gaussian Bernstein function

• Implementing Pareto jump simulation and the Pareto-jump compound Poisson Bernstein function

• Implementing the Exponential-jump compound Poisson Bernstein function

• Add S4 classes for evaluating Bernstein Functions and their higher-order alternating, iterated forward differences

• Provide drop-in wrapper functions to create meaningful distribution parameters

• Refactoring of C++ backend to distribution classes satisfying a multivariate version of the named requirement RandomNumberDistribution

• Refactoring of several internal functions by rewriting them in C++

• Refactoring of simulation algorithms to improve performance

• Refactoring to improve internal representation

• Refactoring of custom assertions

• Improve code coverage, e.g., by adding additional unit tests, an integration test, and statistical unit tests

rmo 0.2

• Re-License under GPL-3 (because of Rcpp dependence)

• Properly handle case rate == 0 in rmo:::sample_cpp (used in rmo_lfm_cpp)

• Properly handle case when compound Poisson process drifts over several barriers during waiting period in rmo:::sample_cpp (used in rmo_lfm_cpp)

• Re-implementation of sampling algorithms using Rcpp

• Use original R-based implementations of simulation algorithms for unit tests

• Improve code coverage for rmo_lfm_cpp (independence case)

rmo 0.1

• Added pure R implementations of various sampling algorithms: rmo_esm, rmo_arnold, rmo_ex_arnold, rmo_lfm_cpp, and rmo_esm_cuadras_auge.

hsloot/rmo documentation built on May 1, 2024, 4:28 a.m.