- preliminary integration of code from
SCnext/mixedSCGLRwritten by Jocelyn Chauvet (issue #11) - moved parallelization to
futureframework. Needfuture.applypackage (issue #16) - added progress to
scglrCrossValandscglrThemeBackwardfunctions usingprogressrpackage (issue #17)
- parameter
nfoldsofscglrCrossValas been renamed tofoldsand now accepts also vectors. For now a warning is issued and provided value is used but update your code as it will likely be removed in a future version. - parameter
mc.coresofscglrCrossValis now deprecated as of the move tofutureframework for parallelization.
- fix bug happening when providing a single Y variable to
sglrTheme(issue #1) - allow to provide folds for cross validation defined by user (issue #2)
- individual beta and gamma for themes were not properly computed (issue #5)
- improve evaluation of linear predictors in
themefunction (issue #8) - remove dependency on
expmpackage (issue #10) - error message for lack of convergence were improved and
scglrCrossValno longer stops when a fold does not converge (issue #12) - add a plot for
scglrCrossVal - fix contrasts error in
scglrTheme(issue #20) - fix some URL in docs
- move from
plsdepot::plsreg2topls::plsrto initialize mixed model components (issue #22) - fix rstudio bug displaying error message when printing a MultivariateFormula (in fact a Formula) (issue #23).
This major version introduces a new feature allowing to group covariates in so called themes.
- added
scglrThemeandscglrThemeBackwardto handle theme oriented selection - reworked
multivariateFormulato handle themes - added new plots targeting themes
- deprecated
barplotin favor ofscreeplot(same parameters)
- Removed LPLS legacy method
- changed from ING to PING
- New method is available : SR (Structural Relevance) see vignette
- Major rewrite of plot styling (not backward compatible)
- Various fixes and improvements (especially when dealing with a single dependant variable)
Initial version of SCGLR
- method LPLS (Local PLS)