Fixed minor bug in
Fixed minor bug with out-of-bag risk for non-cyclical fitting.
Added www and a / after the URLs to make CRAN happy.
Changed http to https to make CRAN happy, hopefully finally.
Minor changes to make CRAN happy.
mboostLSS_fit to avoid writing
(and overwriting) the combined risk in the global environment.
Closes issue #49.
stabsel.mboostLSS for models fitted with FDboostLSS.
Closes issue #51.
Code in ‘man/families.Rd’ failed. Closes issue #53.
predint was not working when multiple variables were included
within a single baselearner.
Closes issue #55.
Fixed CRAN checks regarding
all.equal by adding
check.environment = FALSE.
Export plot functions, streamlined interface of cvrisk and updated manuals. Closes issue #46.
predict.mboostLSS to make betaboost happy.
Updated reference for Thomas et al. Closes issues #47.
Added new fitting algorithm which can be used via the argument
method = "cyclic" is the standard approach which was also
method = "noncyclic" provides a new non-cyclical algorithm.
"noncyclic" method allow faster cross-validation and better results
for stability selection (see below). For details see
Thomas, J., Mayr, A., Bischl, B., Schmid, M., Smith, A., and Hofner, B. (2017),
Gradient boosting for distributional regression - faster tuning and improved
variable selection via noncyclical updates.
Statistics and Computing. Online First.
(Preliminary version: https://arxiv.org/abs/1611.10171).
Stability selection (
stabsel) implemented for
Models can now be fitted with zero steps (i.e., models containing only the offset). This change was mainly realized with the release of mboost 2.8-0. This closes issue #22.
as.families interface to incorporate gamlss distributions
gamboostLSS models was adapted to allow for other link functions.
stabilization = "L2" to use the average L2 norm of
the negative gradient to make the updates for the different distribution
as.families("BI"), which is
actually an mboost family. Closes issue
as.families for 4-parametric
Added Janek Thomas as new author.
Updated ‘inst/CITATION’ due to release of JSS tutorial paper and added new noncyclical fitting paper (also to manuals). The latter closes issue #27.
Updated URL in ‘DESCRIPTION’.
mboostLSS_fit() such that it can be called by
Speed up computations by using
fitted(, type = response) as the latter uses
predict(). Closes issue
Improved output of
summary.mboostLSS(). Closes issue
gamboostLSS_intern() to make FDboost
happy. Caution: Do not use this function.
Fixed tests for stabilization.
Fixed bug in
cvrisk.mboostLSS() which occurred if
families wasn't explicitly specified when fitting the
New interface: Stabilization of negative gradient is now
specified via families, e.g.,
options(gamboostLSS_stab_ngrad = TRUE) is
still supported for backward compatibility but discouraged.
weighted.median which is used in
stabilization (see bugfixes).
Added preliminary version of the tutorial paper as vignette,
which is now available as
package = "gamboostLSS").
summary function (request
Added preliminary version of tutorial paper as reference.
weighted.median to compute MAD when negative
gradients are stabilized.
Fixed handling of gamlss.dist families with
Fixed a bug in
as.families with truncated
(which can be obtained using the package gamlss.tr).
The internal helper
get_qfun is now a generic
Adhere to CRAN policies regarding import of base packages (closes #1).
Changed plain text ‘NEWS’ to ‘inst/NEWS.Rd’
Changes in ‘inst/CITATION’ to make CRAN happy: Citations can now be extracted without the need to install the package.
india data set:
stunting ranged from -600 to 600 but should range from -6 to 6
Fixed unstated dependencies
gamlss.dist:: where necessary
Fixed bug in
predict() returned a vector instead of a matrix
factors were coerced to
added survival to suggests and load it in tests to reflect recent changes in mboost
removed the generic function
risk() as this function is
now defined in mboost
check if response is correct for the
offset is computed
\donttest in examples
added new function
cvrisk for generic
cross-validation (gamboostLSS now requires mboost >=
2.2-3 for the generic function definition)
added a function to obtain cross-validation grids
added new function
as.families to include
gamlss families in gamboostLSS
added new families
GaussianLSS is now the default for
added new functions
plot(predint()) to obtain and plot marginal prediction
added stabilization factor for negative gradients (MAD):
turn on the stabilization by using
options(gamboostLSS_stab_ngrad = TRUE)
added new data sets
lots of minor improvements in the interface and methods
improved manuals in various places
Speed of model fitting was dependent on iteration,
i.e. the algorithm got slower for larger
removed duplicated (and slightly altered) code from
mboost that was required to make gamboostLSS work with
earlier versions of mboost;
gamboostLSS now requires mboost >= 2.0-12
fixed potential problem that might occur if users specify different outcomes for different components
fixed bug in ‘tests/regtest-gamboostLSS.R’
changed the way we use environments
(to get rid of note:
no visible binding)
fixed bug in
moved gamboostLSS to CRAN
some changes to make gamboostLSS work with mboost 2.0.x series
initial development version of gamboostLSS on R-forge