NEWS
dng 1.1
- Added
log = FALSE to both density functions; logarithm remains a
backward-compatible alias and takes precedence when supplied.
- Fixed unsafe vector recycling, empty inputs, invalid parameter handling, and
missing-value propagation.
- Corrected asymmetric split-normal skewness and excess kurtosis, and enforced
the existence conditions for split-t moments.
- Stabilized log-densities, CDF tails, quantile endpoints, and the infinite-df
normal limit.
- Replaced the split-t CDF df-gradient series with integration of the analytic
density score; hypergeometric convergence failures now produce warnings.
- Reduced temporary allocations and switched to direct half-normal/half-t
mixture sampling. Seeded draws differ from earlier releases.
- Added regression tests for asymmetric moments, gradients, tails, and inputs.
dng 1.0.0 (2026-05-03)
- Added
gsplitn() for split-normal CDF and log-density gradients.
- Added and expanded
testthat coverage for split-normal and split-t
distributions, moments, random generation, and gradients.
- Fixed split-t density and kurtosis bugs.
- Modernized package metadata and citation format.
- Kept R documentation generated from C++ roxygen comments.
dng 0.2.1 (2018-06-27)
- Added split-t gradient support through
gsplitt().
- Improved C++ argument recycling with
rep_len().
- Fixed
qsplitn() issues.
- Consolidated distribution and moment documentation.
- Refreshed package metadata, generated Rcpp exports, and tests.
dng 0.2.0 (2018-06-15)
- Added split-normal distribution functions:
dsplitn(), psplitn(),
qsplitn(), and rsplitn().
- Added split-normal moment functions for mean, variance, skewness, and
kurtosis.
- Added split-normal documentation and examples.
dng 0.1.1 (2017-11-22)
- Added split-t distribution functions:
dsplitt(), psplitt(), qsplitt(),
and rsplitt().
- Added split-t moment functions for mean, variance, skewness, and kurtosis.
- Added initial package documentation, citation metadata, and package
infrastructure.
- Improved Rcpp compatibility and performance.