Version 0.23.0: 2015-08-18
- fixed problem with methods "initialize" and "plot"
Version 0.22.0: 2015-07-17
- explicitly import functions from standard packages
Version 0.21.0: 2014-08-28
- new method use.aux.urng() for using a second auxiliary
URNG in Runuran objects that are based on the
acceptence-rejection method.
This makes this kind of genenator suitable for
correlation induction.
- new function set.aux.seed() for seeding the auxiliary URNG.
- moved vignettes from directory inst/doc/ into
directory vignettes/.
- fixed some typos
- internal:
added directory src/scripts with scripts for updating
documentation.
Version 0.20.0: 2012-09-10
- internal:
replaced .Call to internal C function from package 'base'.
Version 0.19.0: 2012-01-31
- pinv.new():
setup is now numerically more robust.
- ud():
added argument 'islog' (with default FALSE).
When TRUE, the log-density is returned (if available).
- new function udmeixner() for Meixner distribution.
- new function udvg() for Variance Gamma distribution.
- new function udgiga():
provides an alternative parametrization for the Generalized
Inverse Gaussian (GIG) distribution.
- udghyp(): implementation of PDF is numerically more stable.
- unuran.details(...,debug=TRUE):
the array of precomputed constants is returned for
methods CSTD and DSTD.
Version 0.18.0: 2011-05-19
- unuran.discr.new(): new argument 'cdf' for adding
CDF to user-defined discrete distribution.
- unuran.details(return.list=TRUE,...):
returned list contains the new field "distr.class" that
contains the type (class) of distribution:
"cont", "discr" or "cmv".
- added more checks in various functions that verify given
arguments.
- removed some obselete lines of C code.
Version 0.17.0: 2011-01-26
- new function unuran.is.inversion():
test whether the generator object implements an inversion
method.
- internal:
* S4 class "unuran" has new slot "inversion" for
storing information about type of generation method
Version 0.16.0: 2010-10-25
- fixed Rd files
- bug fixes
Version 0.15.0: 2010-08-06
- new function up() for computing distribution function
for a unuran object.
- ud(): first argument has been renamed from 'distr' to 'obj'.
argument 'obj' can also be a generator object.
- pinv.new(): when no 'center' is provided by the user, or
when the PDF vanishes or is not finite for the given 'center',
then the method tries to "guess" an appropriate
point for the 'center'.
- special functions (like gamma function) are now taken from
the Rmath library.
Version 0.14.0: 2010-04-29
- new argument 'smooth' for function pinv.new().
- method PINV (pinv.new and pinvd.new) is now more robust
for distributions with "rough" PDFs (e.g. those created by
spline interpolation).
- new function unuran.verify.hat():
Check hat and squeeze of method that implements a
rejection method.
Version 0.13.0: 2010-02-09
- added new function mixt.new() for creating a generator for a
finite mixture of (continuous/discrete) univariate distributions.
Version 0.12.0: 2010-01-04
- unuran.details() now has two new arguments 'show' and
'return.list'. If 'return.list=TRUE' then the function returns
a list that contains some of the data.
- some improved error messages
- fixed bug that caused unuran.details() to hang
Version 0.11.0: 2009-11-17
- method 'r' is defunct.
Please use (the identical but faster function) 'ur'.
- new functions for creating S4 distribution objects for
special distributions:
continuous distributions:
udbeta ... beta
udcauchy ... Cauchy
udchi ... chi
udchisq ... chi-squared
udexp ... exponential
udf ... F
udfrechet ... Frechet
udgamma ... gamma
udghyp ... generalized hyperbolic
udgig ... generalized inverse Gaussian
udgumbel ... Gumbel
udhyperbolic ... hyperbolic
udig ... inverse Gaussian
udlaplace ... Laplace
udlnorm ... log normal
udlogis ... logistic
udlomax ... lomax
udnorm ... normal
udpareto ... Pareto
udpowerexp ... powerexponential
udrayleigh ... Rayleigh
udslash ... Slash
udt ... Student's t
udweibull ... Weibull
discrete distributions:
udbinom ... binomial
udgeom ... geometric
udhyper ... hypergeometric
udlogarithmic ... logarithmic
udnbinom ... negative binomial
udpois ... Poisson
- deprecated functions are defunct.
Please replace
uqhinv by uq
unuran.cmv by unuran.cmv.new
unuran.cont by unuran.cont.new
unuran.discr by unuran.discr.new
urhitro(n,...) by ur(hitro.new(...),n)
urtdr(n,...) by ur(tdr.new(...),n)
Version 0.10.1: 2009-08-06
- bugfix: pinv.new() now also works for R versions compiled with
SunStudio compiler.
Version 0.10: 2009-06-22
- new function 'unuran.packed' for packing 'unuran' objects.
When packed, all information are stored as R lists and allows
handling of 'unuran' objects just as ordinary R objects, in
particular they can then saved and restored in future R sessions,
and they can be copied to other nodes in parallel computing.
- uq() now also supports method 'DGT' (for discrete distributions).
Version 0.9: 2009-02-26
- IMPORTANT:
Paramters 'lb' and 'ub' (lower and upper bound of domain)
do not have default values any more and are obligatory in
all functions with these arguments. Thus thee following
functions have been changed:
o Continuous distributions:
unuran.cont.new
ars.new
itdr.new
pinv.new
srou.new
tdr.new
If R stops when calling these functions with the error
message "domain ('lb','ub') missing", then one has to add
the formerly defaults "lb=-Inf" and/or "ub=Inf" for the
missing arguments.
o Discrete distributions:
unuran.discr.new
dari.new
If R stops when calling these functions with the error
message "domain ('lb','ub') missing", then one has to add
the formerly defaults "lb=0" and "ub=Inf" for the missing
arguments.
We observed a possible pitfall with the former interface.
When a user wanted to draw a sample from a distribution with
positive domain but forgot to set "lb=0" then it was
possible that the generator (silently!) did not work as
expected and caused invalid simulation results.
Forcing to set the domain reduces this risk.
Thus we decided to stay on the safe side.
We appologize for this inconvenience.
- fixed syntax errors and inconsistent formating in Rd files.
Version 0.8: 2008-11-26
- reorganized functions into three groups:
* Runuran-Distribution:
easy-to-use sampling algorithms for particular
distributions.
* Runuran-Universal:
interface to a carefully selected collection of UNU.RAN
methods with their most important variants.
* Runuran-Advanced:
wrapper to the UNU.RAN string API.
- improved documentation,
added vignette
- new function uq() for computing quantiles approximately.
- new functions for directly creating generator objects
for selected UNU.RAN methods:
ars.new()
dari.new()
dau.new()
dgt.new()
hitro.new()
itdr.new()
pinv.new()
srou.new()
tdr.new()
vnrou.new()
- new functions for creating UNU.RAN distribution
objects:
unuran.cmv.new(...) replaces unuran.cmv()
unuran.cont.new(...) replaces unuran.cont()
unuran.discr.new(...) replaces unuran.discr()
Remark: The default for argument 'islog' has been changed to
'islog=FALSE'.
- deprecated functions:
uqhinv()
urdau()
urdgt()
urhitro()
urtdr()
- contains UNU.RAN version 1.3.0
Version 0.7: 2008-04-21
- print() now shows some information about a generator
object.
- new function unuran.details() for printing detailed
information about a generator object.
- new function uqhinv() for compting quantiles approximately
using UNU.RAN method 'HINV'.
Version 0.6: 2007-12-10
- added slots for a rectangular domain and the center
(point in main region of distribution) for multivariate
distribution class 'distr.cmv'.
- new method 'MVTDR' in UNU.RAN.
- fixed broken optional arguments in routines urtdr() and
urhitro().
(The optional arguments were not passed to the given
density function in contradiction to the help page of
these routines.)
Version 0.5: 2007-08-01
- added S4 classes for distribution objects:
This allows to define distributions by means of
R functions instead of using the UNU.RAN string API
(which is still supported).
- added support for multivariate distributions.
- added routine 'urtdr'.
Version 0.4: 2007-05-16
- First public release.