2007-02-25 David Luethi
ghyp V0.8.3
* 'ghyp' released.
2007-03-26 David Luethi
ghyp V0.9.0
* Bug fixed in lin.transf.
* Slots aic and fitted.params added to mle.ghypuv and mle.ghypmv.
* Function stepAIC.ghyp added.
* Argument names "log" and "legend" changed to "log.hist" and
"plot.legend" respectively in qqghyp and hist.
* Some corrections in the vignette.
2007-06-08 David Luethi
ghyp V0.9.2
* Internal function abar2chipsi: epsilon changed to
.Machine$double.eps.
* Internal function internal.dghyp, internal.dghypmv: Warning
message changed.
* Internal function mle.default, fit.ghypmv: Output properly
formatted with sprintf.
* Spline interpolation in internal function internal.dghyp for
transition to singularity for the Variance Gamma case.
* Generic functions logLik and AIC added.
* Some changes in the package vignette.
2007-07-05 David Luethi
ghyp V0.9.3
* Some changes in the package vignette and documentation.
* Function ghyp.moment added.
2007-09-14 David Luethi
ghyp V1.0.0
* fit.ghypmv: default relative tolerance set to 1e-8.
* fit.ghypmv and fit.ghypuv: Parameter transformation changed to
"log".
* t.transform changed.
* inf.t.transform: added; inverse of t.transform.
* Some changes in the package vignette and documentation.
* Bug fixed in check.opt.pars: Check for 'symmetric' and
opt.pars['lambda'] and set opt.pars['lambda'] to FALSE when
'symmetric' is TRUE.
* hist.ghypuv: Warnings now more meaningful.
* check.norm.pars: Warnings now more meaningful.
* Function 'lin.transf' replaced by the generic 'transform'.
* Function 'redim' replaced by the generic '[.ghyp'.
* Function 'ghyp.param' replaced by the generic 'coef'.
* Methods 'show' changed and 'summary' added for objects of class
'mle.ghyp'.
* Argument 'standardize' introduced in the fitting routines
'fit.ghyp..'
* Methods 'plot' and 'lines' added.
* Initial parameter value 'nu' changed in 'student.t', 'fit.tuv',
'fit.tmv'.
2008-02-05 David Luethi
ghyp V1.1.0
* fit.ghypuv, fit.ghypmv now equipped with a rather reasonable
error handling.
* Function 'ghyp.name' added.
* Internal function besselM3: For nu == 0.5 in besselK there is a
faster expression for the bessel function which is now
implemented.
* Some minor changes in the vignette.
2008-03-07 David Luethi
ghyp V1.2.0
* Special case of a Gaussian distribution implemented (constructor
'gauss' added).
* Deprecated functions 'lin.transf', 'ghyp.params' and 'redim'
removed and replaced with 'transform', 'coef' and '[',
respectively.
* ghyp constructors added for 'alpha.beta' parameters.
* Vignette updated and newly implemented parametrization
documented.
* Error handling improved for qghyp, pghyp, (ESghyp), qgig,
pgig, (ESgig).
* Constructor 'ghyp' modified such that the variance of the
mixing-variable is not required for the symmetric case.
* Student-t density now allows "chi.psi" and "alpha.delta"
parametrization.
* dghyp and pghyp now handle NA's, NaN's and Inf's properly.
* Some improvements in the manual.
* Internal functions 'is.gaussian', 'is.student.t' and
'is.univariate' added.
* Internal functions 'get.stepAIC.ghyp' added.
* Univariate symmetric Student-t distribution now use R built-in
functions.
2008-07-07 David Luethi
ghyp V1.3.0
* Option lower.tail added in pghyp.
* fit.ghypuv: Initial values of mu and sigma are now median and
mad, respectively (instead of mean and sd).
* ghyp-constructors: Default parameter values changed for skewness
parameters gamma and beta to rep(0, length(mu)).
For dispersion matrices sigma and Delta to
diag(rep(1, length(mu))).
* The function 'qqghyp' now has an argument 'add' which determines
whether the points should be added to an existing plot window.
* Constructor function "ghyp": all(gamma) == 0 changed to
all(gamma == 0).
* Internal function 'is.symmetric.t' changed to
'is.student.t(object, symmetric = NULL)'.
* 'coef' applied to a Student-t distribution now also returns the
degree-of-freedom 'nu', which is nu = -2 * lambda. Function
'show.ghyp' adapted accordingly.
* It is now possible to switch from the 'chi.psi' parametrization
to the 'alpha.bar' parametrization.
* Package vignette enhanced.
* Slot 'trace.pars' added to class 'mle.ghyp'. This slot contains
a list (optionally in case of a multivariate fit) which stores the
parameter values during the fitting procedure.
* 'fit.ghypmv' is equipped with a boolean 'trace' as a consequence
of the above modification. In the univariate case 'trace' is
always true, meaning that the parameters will be stored in the
slot 'trace.pars' in any case.
* 'ghyp.fit.info' returns a formatted version of the data in the
slot 'trace.pars'.
* Initial value of 'alpha.bar' changed to 0.5 instead of 0.1 in
'fit.ghypuv'.
* Function 'ghyp.omega' added. 'ESghyp' and 'ghyp.omega' are now
described in the .Rd file 'ghyp-risk-performance'.
2008-10-28 David Luethi
ghyp V1.4.0
* 'ghyp.moment': in case of central moments: wrongly took
parameter 'mu' instead of the expected value in case 'central' was
TRUE. 'ghyp.moment' was extended such that it takes analytical
expressions and avoids numerical integration whenever possible.
* The package 'ghyp' now depends on R >= 2.7. Function 'identity'
was removed accordingly.
* Functions 'ghyp.skewness' and 'ghyp.kurtosis' added.
* 'ghyp' is on R-forge.
* Instead of putting each function in a different file, functions
are now collected in 8 files: 'ghypClasses', 'ghypConstructors',
'ghypFitting', 'ghypGenericMethods', 'ghypGIG', 'ghypInternals',
'ghypMethods', 'ghypWrappers'.
* Internal functions now all come with a '.' in front of the name.
* Internal function 'internal.dghyp' and 'internal.dghypmv'
renamed to '.dghypuv' and '.dghypmv' respectively.
* Package vignette extended.
* Function 'ghyp.dim' added. Returns the dimension of a 'ghyp'
object.
* Internal function '.is.symmetric' added.
* Dataset 'indices', representing 5 financial indices,
and documentation file 'indices.Rd' added.
* Function 'portfolio.optimize' extended for tangecy and global
minimum risk portfolios. Some names of the arguments had to be
changed.
2008-10-30 David Luethi
ghyp V1.5.0
* 'transform.ghyp' changed: If data is stored in slot 'data' of an
object inheriting from class 'ghyp', 'data' will be transformed as
well.
* Generic function 'scale' added to conveniently center and/or
scale a 'ghyp' object to expectation zero and/or unit variance.
* Function 'rgig' modified. It now uses the random number
generator for the GIG distribution from the Rmetrics package
'fBasics'.
2009-05-20 David Luethi
ghyp V1.5.1
* 'student.t' constructor and internals modified such that it
works for non-finite variance and Cauchy distribution in
particular.
* Documentation and Vignette improved...
2009-10-11 David Luethi
ghyp V1.5.2
* Documentation and Vignette improved...
* 'qghyp': In case of student-t convert obj@sigma to 'numeric'
from 'matrix'.
* Constructors 'ghyp', 'NIG', 'hyp', 'VG', 'student-t' now accept
matrices for 'mu' and 'gamma' and coerce them to vectors. 'Gauss'
accepts a matrix for 'mu'.
* Warning issues for Rd-files under R-2.11.0 resolved.
2010-10-11 David Luethi
ghyp V1.5.3
* Vignette: Typo in eq. (5.8) removed in courtesy of Simon Keel.
* rgig: GIG random number generation is now done in C with a
significant increase in performance. The code was kindly provided
by Ester Pantaleo and Robert B. Gramacy.
2010-11-29 David Luethi
ghyp V1.5.4
* In NAMESPACE: S3Method changed to S3method as this will become
case-sensitive in R-2.13.0.
2011-03-30 David Luethi
ghyp V1.5.5
* Non-ASCII characters removed from the vignette.
2013-02-04 David Luethi
ghyp V1.5.6
* lik.ratio.test: pchisq(q, df, lower.tail = FALSE) instead of 1 -
pchisq(q, df, lower.tail = TRUE). Leads to increased numerical
stability. In courtesy of Saeed Ahmadi.