2.6.4
Export fit.fkml.moment
Improve error message in rgl() for invalid parameter values
na.rm now works for starship method in fit.gpd()
Improve docs
2.6.3
New email address and URL
plot.starship documentation correctly describes working with fit.fkml
fit.gpd() now allows method="starship", then calls starship()
2.6.2
gl.check.lambda bugfix
2.6.1
Add LambdaZeroEpsilon argument to fit.gpd to allow wider range of Lambda values to count as zero, and thus Quantile Based Skew Logistic. This is necessary because for abs(lambda) < 1e-15, SE(lambda) can be over 5e5
Correct gl.check.lambda, which was missing correct parameter values in regions 1 and 2 of the Ramberg & Schmeiser type
2.6
plot method for starship, plot.starship now defaults to one.page=FALSE
The fit.fkml function now;
* Returns and prints trimming values in Trimmed mean estimates
* Has the LMom method which is trimmed mean with zero trimming
* Has MoM, method of moments
Error in documentation for fit.gpd fixed
2.5
Standard Errors for the L Moment estimates for the GPD type
fit.gpd defaults to recording cpu time usage (so that it is the
same as fit.fkml)
2.4.2
Fix bug in fit.gpd
BetaLambdaLambda function added (utility function for gpd L-Moment estimation)
2.4.1
Fix pgl() example comment - inverse.eps had defaulted to .Machine$double.eps since version 1.9
2.4
Method of L Moment estimates for the gpd type
2.3.3
print.starship, summary.starship and plot.starship now display
estimation method and gld type
2.3.2
optimisation in fkml.moments.val is now constrained to the region on lambda3, lambda4 space where moments exist.
Documentation for moment fitting notes multiple parameter sets with the same first 4 moments
2.3.1
Bugfix - plotgld() was giving warnings due to use of the deprecated qdgl
2.3
Added calculation of moments for FKML given parameters and fitting
FKML with moments (given moment values - not calculating the moments yet)
Function qdgl deprecated and replaced with dqgl as it actually calculates the density quantile function
2.2.1
Bugfixes - DLA wasn't being called correctly
Also includes some C changes for fit.fkml. The C code now assumes the data to be fitted has already been sorted
2.2
New estimation methods for the FKML parameterisation:
Numerical MLE, Maximum Spacings Product, Titteringtons, L Moments,
TL Moments, Distributional Least Absolutes
Starship has C speedups from Ben Dean, in fit.fkml function
fit.fkml is called by starship() if param="fkml"
Bugfixes - parameter checking error messages
2.1
C speedups from Ben Dean (internal version)
2.0.1
removed \usage lines wider than 90 characters
2.0
Added VSK/GPD parameterisation (all basic functions and starship)
Removed default values of initgrid in starship.adaptivegrid() - users now
need to specify an initial grid if using this function. starship()
still provides the same default when starship.adaptivegrid is called.
C error message improvements
Changes to the way the default initgrid is used in the starship function
1.9.4
Debugged
.First.lib removed
stderr references removed
1.9.3
Debug plotgld() for pathological densities
1.9.2
NAMESPACE added
1.9.1
Bugfix: plotgld was wrongly labelling some densities [such as
c(0,1,4,5),param="rs"] as pathological, which should be restricted to those
like c(0, 1, 0.03, 0.7), param = "rs".
Bugfix to plot.starship code and documentation
1.9
Provided qdgl() as a duplicate of dqgl() because these are actually the quantile density functions, not the density quantile functions
Allowed for the FKML parameterisation to be referred to as either FKML or FMKL to allow for the different uses in the literature.
Added a plot method for the starship class. This compares the fitted gld to the data.
Added add argument to plotgld() and plotglc()
Changed inverse.eps default to .Machine$double.eps
1.8.5
Added QQ plot to compare two sets of parameter values
1.8.4
Added "starship" class, with methods for print() and summary()
Documentation fixes
1.8.3
Fixed bug stopping functions working if the parameter values are extracted from a dataframe
Added ability to set inverse.eps in starship() - default is still 1e-08
really fixed dots.end bug in plotgld
1.8.2
Fixed dots.end bug in plotgld
Changed exit() to error() in C code
1.8.1
plotgl warning (1.8 change) bugfix
Removed the infinite parameter values check
1.8
plotgl warns about pathological densities and plots slightly differently
where the theoretical (finite) minimum is actually a long way
from Q(.Machine$double.eps) or similar with the maximum
Bugfix: DOS CR characters in R file causing problem on alpha
1.7.2
Bugfix: no visible binding for param in .qdgl.fmkl and friends
1.7.1
Bugfixes: initgrid default value for RS parameterisation
Documentation fix starship
1.7
Added support for the 5 parameter FMKL version (FM5 parameterisation)
Changed syntax for starship.adaptivegrid
Allowed parameters to be given as a vector, rather than with individual
arguments
Fixed bug which could have slowed F(x) calculation for pgl(...,param="fmkl")
when lambda4=0, lambda3!=0
1.6
Documentation Example fixes
1.5
Fixed "assign long to int" bug in C code
1.4
Documentation fixes
Correction of parscale in optim section of starship()
1.3
Added qqgl - make QQ plots for data against specified gld
1.2
INDEX file correction
1.1
Documentation corrections
1.0.1
Permissions correction
1.0.0
starship estimation, using grid or grid and optim
Range checking
Corrections from Steve Su
0.9
Corrections from David Scott
Added plotgld
0.8
improved error messages
Fixed bug in qgl and qdgl for parameterisation RS (was using FMKL
parameterisation values even if RS was requested)
added alias for GeneralisedLambdaDistribution to
GeneralisedLambdaDistribution.Rd as requested by Kurt
0.7
Note that R CMD build complains about the INDEX file, but I want it the way I
have it
R CMD check changes to .Rd files
0.6
URL changes to Newcastle
0.5
Took the mac makefile out again
0.4
Typo correction in man/
Added mac makefile
Added Maintainer field
Corrections to DESCRIPTION
0.3
First released to CRAN
0.2
0.1
Internal versions