Disabled OPENMP compilation flags due to potential test failures on some platforms, e.g. Solaris.
POUMM 2.1.5 2019-03-27
Improved documentation for some functions (specifyPOUMM).
Fixed a bug: duplicated rows for “g0” in the summary of a POUMM object.
Compatibility with v1.12.2 of package data.table - using rep instead of recycling in columns with more than 1 but fewer than the max-number of rows (see https://github.com/venelin/POUMM/pull/2).
POUMM 2.1.3 Unreleased
Improved man page for gPOUMM function.
Testing minor changes in the source package - store RData files in lfs.
Testing github release and zanodo doi-issuing.
POUMM 2.1.2 2018-11-15
Temporarily removed bivariate posterior density plotting functionality; univariate posterior density plotting still available.
Removed dependency to RcppArmadillo
Fixed a bug in the C++ likelihood calculation for the case when both sigmae = 0 and se = 0.
Refactored unit tests.
Enabled test-coverage report with inserted badge in the README.Rmd.
Enabled travis continuous integration with a status badge in the README.Rmd.
Some bugs were fixed.
POUMM 2.1.1 Unreleased
Release available from github only
POUMM 2.1.0 Unreleased
SPLiTTree dependency is renamed to SPLITT
Release available from github only
POUMM 2.0.0 Unreleased
Using SPLiTTree for the pruning likelihood calculation.
Automatic web-site generation using the R-package pkgdown.
Fixing test-errors on r-patched-solaris-x86 and notes on r-devel-linux-x86_64-fedora-clang.
Fixed memory errors reported by ASAN and valgrind.
Improved default values for the parameter sigmae.
POUMM 1.3 Unreleased
Parallel likelihood calculation: (tested on Linux using Intel complier v16.0.0).
Added possibility to specify known measurement standard error for each tip of the tree. (see ?POUMM).
Added a bivariate densplot of the posterior samples displaying a pairs-plot, i.e. a matrix of plots with marginal densities at the diagonal, scatter plots on the lower triangle and correlation on the upper (see ?plot.summary.POUMM); possibility to change the color-palette used in the plot with color-blindness friendly default colors.
New default parameter settings: now the ML-limits (parUpper and parLower) as well as the priors for the parameters alpha and sigma are scaled according to the length of the tree (tMean) and the empirical variance of the trait (zVar) (see ?specifyPOUMM). Please note that the default parameter settings may change in future releases.
Improved ML-fit: now the optimization is run 50 times starting from different points in the search region. The ML-fit is still kept deterministic, i.e. consecutive runs should find the same optimum.
Fixed a dependency on the adaptMCMC package (now added as an import).