Added explicit atomic flag for he() in TMB since usingAtomics() in
default argument of function is not picked up in some O/S.
tsdistributions 1.0.2
Replaced PI with M_PI to pass strict header checks.
Fixed missing package link for bkde function in documentation.
Translated the distribution c functions to Rcpp.
Removed zoo dependency (not needed).
tsdistributions 1.0.1
Added the semi-parametric piece-wise distribution
(spd). As this is a special type of distribution, it gets
its own class for specification and estimation. Added p, d, q, r
functions and other methods (summary, vcov, coef etc).
Fixed the noRemap Additional Issue on CRAN checks (renames error to
Rf_error in C++ code).
Added a unit test for the spd in the test-quantile.R file.
Added an html vignette for the spd.
tsdistributions 1.0.0
Initial CRAN submission.
Cleaned up c-code naming of distributions and removed redundant
code.
Removed gig.c and nig.c which were used for generating random
numbers. Now using imported functionality (rghyp) from
GeneralizedHyperbolic package with transformed parameters.
Cleaned up documentation and naming convention for file names with
methods.
Renamed Generalized Hyperbolic acronym to ‘gh’ (from ‘ghyp’).
Added a tsprofile method.
Added an html estimation example vignette.
Added an html profile example vignette.
Added a pdf vignette for location scale distributions outlining the
formulation of each distribution.
tsdistributions 0.3.3
Fix to the Generalized Hyperbolic Distribution estimation code as
well as the scaled besselK function in the presence of fixed lambda. The
fix somewhat negates the speedup from the previous version when lambda
is fixed, but ensures correctness.
tsdistributions 0.3.2
The Jacobian, which is used in the calculation of the QMLE and OPG
covariance matrix is now calculated using TMB rather than numDeriv since
that had problems for the ghst and gh distribution giving hugely
erroneous results. The method of calculating it loops through each data
point, and is rather hacky (but works). Using autodiff::jacobian
directly in the TMB C++ code is proving challenging.
Removed numDeriv dependency
tsdistributions 0.3.1
Fix to Fernandez and Steel skewed distributions quantile function
(qsnorm, qsged and qsstd).
Added additional unit tests.
tsdistributions 0.2.0
Fix to snorm distribution (dsnorm) for cases when variable z == 0.
Other skewed distributions used signum function but this distribution
seems to have missed the 0 case for some reason.
Added vcov methods and other sandwich estimators (QMLE, OP,
NW).
Added BIC and AIC and created a print method for the summary.