0.38.196 (Oct 18, 2019)
* An issue in the internal C code to parameter estimation performed by heavyLme
was fixed (thanks to Manuel Galea for reporting this bug).
* Added a new function heavyGrubbs for estimation in the Grubbs model using
thick-tailed distributions (see Osorio, Paula and Galea 2009, CSDA 53, 1249-1263,
doi: 10.1016/j.csda.2008.10.034).
* Added thermocouples dataset.
0.38.19 (Feb 13, 2018)
* Now underlying code for location-Scatter estimation is based on an online
algorithm (previously a 2-pass algorithm was used).
* Added routine to compute the 1st and 2nd derivatives of the regularized
incomplete gamma function.
0.38.1 (Apr 5, 2017)
* Added routines to evaluate density, cumulative distribution, quantile functions
and generation of random deviates from the right truncated gamma distribution.
0.38 (Jul 7, 2016)
* Fixed bug on the output of heavyLm in the case of multivariate responses
with only one regressor variable (thanks to Maria de los Angeles Villena
for reporting this issue).
* Corrected error in estimation procedure in heavyLme when the number of
observations per subject (group) is insuficient to obtain the estimators
(pointed out by Ailis Mone, Joffrey Jouma'a and Juvencio Nobre).
* Added fitter functions heavyLm.fit and heavyMLm.fit for linear regression
models.
* Recoded functions for matrix computations using LAPACK entry points.
0.3 (Mar 21, 2016)
* Feature added to heavyLm function in order to handle multivariate linear
regression models.
* Corrected error in computation of penalty term of P-splines smoothing.
* C code of Brent's method for unidimensional minimization was recrafted
from 'localmin' ALGOL procedure described in Brent (1973), Algorithms
for Minimization Without Derivatives, Dover, New York, pp. 79-80 and
188-190.
* Changes to NAMESPACE file to compliance with new guidelines described
in 'Writing R Extensions' Section 1.5.
* Added dialyzer dataset and examples.
0.2-35 (Sep 1, 2014)
* Added heavyFit function for the estimation of the location and scatter
matrix using multivariate heavy-tailed distributions
* Added S3 print and summary methods for heavyFit objects
* Added creatinine dataset and examples
0.2-3 (Mar 22, 2013)
* Added a new function heavyPS for smoothing using P-splines considering
thick-tailed distributions (as described in Osorio 2016, AISM 68, 589-619,
doi: 10.1007/s10463-015-0506-0)
* Added S3 methods for print and summary for heavyPS objects
* Added estimation of the shape parameters for the Student-t and slash
distributions
* Unidimensional optimizer required for the shape parameter estimation use
the Brent's method taken from extracted from R-2.15.3 (since R 3.0.0
several non-API entry points like Brent_fmin, see appl/fmin.c file, are
no longer available)
* Added functions for multivariate random number generation considering
gaussian (normal), Student-t, slash, contaminated normal and the uniform
distribution on the p-dimensional sphere distributions (underlying code
was written in C to ensure a reasonable computational speed)
* Added life dataset and examples
* Removed l1fit function (which since April 4, 2015 this function is available
in L1pack) and housePrice dataset
* Now QR decomposition/operations and SVD decomposition are based in Linpack
entry points
0.2 (Jan 3, 2011)
* Recoded linear algebra operations in order to use the LAPACK entry points
available from R
* heavy.lme function was renamed as heavyLme
* Added heavyLm function for estimation in (univariate) linear regression
* Added l1fit function for L1 estimation in linear regression (using a
slight modification of Barrodale and Roberts Fortran code)
* Fixed error when computing fitted values and residuals by heavyLme
function
* Added S3 print and summary methods for heavyLm and heavyLme
* Added cyt, ereturns and housePrices data and examples
* Added information about citation of heavy package
* Several code cleanups, reduced data movement in some matrix computations
0.1-1 (Dec 15, 2009)
* Initial release of the heavy package on CRAN