2011-05-12 Berwin A Turlach * DESCRIPTION (Version): New version is 1.2-12 * COPYRIGHT: Changed license to GPL >= 2 to conform with information in DESCRIPTION. * Changed e-mail address to Gmail e-mail in various files. 2010-04-12 Berwin A Turlach * DESCRIPTION (Version): New version is 1.2-11 * DESCRIPTION (Maintainer): Changed e-mail address from former NUS e-mail to UWA e-mail. * man/gl1ce.Rd: updated link from [base]{glm.control} to [stats]{glm.control} 2009-01-26 Berwin A Turlach * Added check in l1ce() and gl1ce() whether matrix passed to C routine contains only constant column to avoid attempts for fitting non-sensical models. * Corrected errors in gl1ce.object.Rd and l1celist.object.Rd found by now Rd parser of (future) R-2.9.0. 2008-09-11 Berwin A Turlach * Removed the Makefile from inst/doc since I found a "pure" LaTeX way to check for missing PDFs that are to be included with \includepdf. 2008-09-08 Berwin A Turlach * Added a Makefile to inst/doc since "R CMD check" does not behave in documented way, it actually tries to build the vignette PDF. 2008-09-07 Berwin A Turlach * Various .Rd files contained an unmatched right brace. Fixed those files. Reported by Tim Hesterberg . 2007-09-30 Berwin A Turlach * Updated various .Rd files to cope with changes in R-2.6.0 * Updated some .q files to cope with changes in R-2.6.0 2006-01-06 Berwin A Turlach * DESCRIPTION (Version, Date, Maintainer) * man/predict.gl1ce.Rd: cope with change to all.equal() in R-2.3.x 2005-06-10 Martin Maechler * DESCRIPTION (Version, Date): 1.2-2 * src/lasso.c (lasso): indent and use spaces "properly". * R/gl1ce.q (gl1ce): use eval(m, parent.frame()) * R/l1ce.q (l1ce): instead of eval(m, data) * R/l1ce.q (l1ce): warn about rank only when p <= n 2005-02-11 Martin Maechler * R/l1ce.q (l1ce): improve error message for 'constant column' 2004-04-13 Martin Maechler * DESCRIPTION (Version, Date): 1.2-0 * R/l1ce.q: fix intercepts for multi-bound situation (reported by Martin Keller-Ressel). * tests/l1ce-tst.Rout.save: modified correspondingly and for : * tests/l1ce-tst.R: new consistency test (for the above problem) 2004-02-05 Martin Maechler * DESCRIPTION (Version): 1.1-0 * R/summary.q (summary.gl1ce): $coefficients should be "usual" matrix, not a 1-dim array. {The docs do not talk about it.} * R/print.q (print.summary.gl1ce): coefficients; also visualize '0'! * R/predict.q (predict.gl1ce): type = "response": fix R port * R/gl1ce.q (family.gl1ce): new method * man/gl1ce.Rd: and docu * man/predict.gl1ce.Rd: add \examples{} * man/residuals.gl1ce.Rd: \seealso{} + cosmetic 2004-02-05 Berwin Turlach * R/gl1ce.q (gl1ce): bug fixes (3 places), particularly fitted values. 2003-12-29 Martin Maechler * DESCRIPTION (Version): 1.0-4 * src/lasso.c : get rid of 3 minor compiler warnings * man/coef.l1ce.Rd: do not \link{coef.default} 2003-12-15 Martin Maechler * R/formula.util.q (is.gEnv): new function, needed for testing about .GlobalEnv {gave wrong warnings}. * tests/l1ce-tst.R: RNGversion("1.6.0") added for back compatibility * man/predict.l1ce.Rd: now the newdata works, add example * man/summary.l1ce.Rd: usage for print.summary.l1ce * man/...: a few more such changes * R/gl1ce.q (gl1ce): PACKAGE = "lasso2" was missing * R/predict.q (predict.l1ce): use "term.labels", not "variables" (fix by Berwin himself!). 2003-07-09 Martin Maechler * R/formula.util.q (merge.formula): keep formula environment (and warn if differing). * man/merge.formula.Rd: add example 2003-03-22 Martin Maechler * R/vcov.q (vcov): generic only for older R versions 2002-04-09 Martin Maechler * release 1.0-2 to CRAN * man/l1ce.Rd: correct reference to JCGS paper * man/gl1ce.Rd: link to Berwin's website mentioning Justin Lokhorst's honors report. * inst/doc/justin.lokhorst.ps: is now removed * inst/doc/Manual-wo-help.pdf: instead of ps. * man/qr.rtr.inv.Rd: new file documenting qr.rtr.inv() 2002-02-25 Martin Maechler * man/zaux.Rd and zaux.l1celist.Rd : renamed from aux.* since that is `invalid' on Windows. 2002-02-08 Martin Maechler * DESCRIPTION (Version): 1.0-2 (not submitted; for internal use) * tests/l1ce-tst.R: new file; first testing, also n < p * R/misc.q (aux.l1celist): simplification * R/l1ce.q (l1ce): patch to improve working with rank deficient X e.g. (n < p)! * Changelog: Appended Berwin's S-plus original one. * man/lasso-internal.Rd: added (to get rid of last warning). * man/plot.l1celist.Rd: * R/plot.q: use type = "b" and other more useful default arguments * R/force_ld.q is garbage for R --> rm! 2002-02-06 Martin Maechler * First release to CRAN --------------------- * DESCRIPTION (Version): 1.0-1 (as Berwin explained, "2" is no version number!) * src/lasso.h: define macro Sprintf() = Rprintf() in R, else = printf() * src/lasso.c: s/printf/Sprintf/ * R/gl1ce.q (gl1ce): replace as.family; use is.R() for R differences (such that this still runs in S+ as before!). s/check.vector/Cvector/ {readability}. Many more (using R's glm.fit()) -- now +/- working * R/summary.q (summary.gl1ce): R adaptions * R/print.q (print.summary.gl1ce): R adaptions * man/deviance.l1ce.Rd: and almost all (!!) others: added \keyword{}; added quite a few hyper \link{} s; sometimes merged the almost identical *.l1ce and *.l1celist files; added missing argument lists; .... * R/*.q: generic and methods *must* have maching argument lists! 2002-02-05 Martin Maechler * pre-version (Version): 2.1-0 made available for Berwin to look at. * MM-NOTES: 1) - 7) explaining open questions 2002-01-26 Martin Maechler * DESCRIPTION: a first unreleased version at least runs R CMD check to completion; still many warnings * R/l1ce.q (l1ce): cosmetic; added qr.rtr.inv() {and a note on why I think one should change..} fixed two severe `typos' : 1) single-bound && something.to.sweep.out ("orig" shows) 2) multibound case: `fit' was used before defined ----------------------------------------------------------------------------- ^^^^^^^^^ R Package ----------------------------------------------------------------------------- ----------------------------------------------------------------------------- vvvvvvvvv S-Plus library section ----------------------------------------------------------------------------- Wed Dec 8 23:35:43 1999 Berwin Turlach * First.lib.template: Changed to new release * predict.q: Modified predict.l1ce to support offset for predict.gl1ce. Mimics predict.{lm,glm} behaviour. * l1ce.q, gl1ce.q: Changed some t(foo)%*%bar to crossprod(foo,bar). Fixed a bug that appeared while playing around with update (subscription error when putting end result together). Tue Nov 30 06:31:53 1999 Berwin Turlach * summary.q: Added glm stuff. * summary.gl1ce.d: Initial revision * residuals.q: Added glm stuff. * residuals.gl1ce.d, print.summary.gl1ce.d: Initial revision * print.q: Added glm stuff. * print.gl1ce.d: Initial revision * predict.q: Added glm stuff. * gl1ce.d, gl1ce.object.d, gl1ce.q, predict.gl1ce.d: Initial revision * deviance.q: Added glm stuff. * deviance.gl1ce.d: Initial revision Tue Nov 30 06:15:53 1999 Berwin Turlach * vcov.q, residuals.q: Modified to make port to R easier. * plot.q: Added parameter `constrained'. Modified to make port to R easier. * plot.l1celist.d: Added documentation for additional parameter. * misc.q: Added function `tr'. Modified to make port to R easier. * l1ce.q, gcv.q, fitted.q, deviance.q: Modified to make port to R easier. * coef.q: Added parameter `constrained'. Modified to make port to R easier. * coef.l1celist.d, coef.l1ce.d: Added documentation for additional parameter. * First.lib.template: Changed way of dynamic loading. Fri Dec 18 06:42:48 1998 Berwin Turlach * misc.q: Added tr() command. (Arman Maghbouleh pointed out that it was missing and is not in standard S-Plus.) * tr.d: Initial revision Wed Nov 4 07:50:55 1998 Berwin Turlach * First.lib.template: Changed release number to 1.2 and saved as snapshot. * Finished the documentation and the manual. Sun Sep 27 02:02:02 1998 Berwin Turlach * summary.l1ce.d, l1ce.d, Prostate.d, Prostate.q: Corrected spelling of prostate. Thu Sep 24 10:30:02 1998 Berwin Turlach * First.lib.template: Changed to Release 1.1 and saved as snapshot. * l1ce.q: If the object that is given down to "data" is not a data frame we now try to coerce it into one. Wed Sep 23 10:22:23 1998 Berwin Turlach * l1ce.q: Fixed a bug in l1ce in the case that a vector of bounds is given (coefficients of swept out variables were not handled correctly). Sun Sep 20 09:44:18 1998 Berwin Turlach * First.lib.template: Changed to Release 1.0 and saved as snapshot. * summary.l1ce.d, l1celist.object.d, l1ce.object.d, l1ce.d: Initial revision Fri Sep 18 11:32:58 1998 Berwin Turlach * First.lib.template: Changed to Release 0.12 and saved as snapshot. * [[.l1celist.d, aux.l1celist.d, aux.d, plot.l1celist.d, vcov.l1celist.d, vcov.l1ce.d, vcov.d, residuals.l1celist.d, residuals.l1ce.d, print.summary.l1ce.d, print.l1celist.d, print.l1ce.d, predict.l1ce.d, gcv.l1celist.d, gcv.l1ce.d, gcv.d, formula.l1celist.d, formula.l1ce.d, formula.q, labels.l1celist.d, labels.l1ce.d, labels.q, Prostrate.d, Prostrate.q, Iowa.q, Iowa.d, merge.formula.d, is.formula.d, fitted.l1celist.d, fitted.l1ce.d, deviance.l1celist.d, deviance.l1ce.d, coef.l1celist.d, coef.l1ce.d: Initial revision Tue Sep 15 11:23:42 1998 Berwin Turlach * First.lib.template: Changed to Release 0.11 and saved as snapshot. * l1ce.q: The `contrasts' and `assign' result from model.matrix are now returned as components of l1ce objects and attributes of l1celist objects. (Analogous to what is found in lm.fit). Sun Sep 13 04:08:42 1998 Berwin Turlach * First.lib.template: Changed to Release 0.10 and saved as snapshot. * vcov.q: Added check in `opt.covmat' whether `bound==0'. * summary.q: Programmed and tested `summary.l1ce'. * print.q: Added and tested `print.summary.l1ce'. * gcv.q: Made `gcv' a generic method. Programmed and tested `gcv.l1ce' and `gcv.l1celist'. * summary.q: Initial revision Fri Sep 11 08:55:26 1998 Berwin Turlach * gcv.q: Initial revision Thu Sep 10 10:23:10 1998 Berwin Turlach * First.lib.template: Changed to Release 0.9 and saved as snapshot. * l1ce.q: The terms in the `formula' are calculated and given back too. We need those in the predict function. Minor cosmetic changes to error/warning messages. * misc.q: Corrected bug in transfer from attributes of objects of class l1celist to components of objects of class l1ce. Streamlines the code and works also if new attributes/components are added. * predict.q: Adapted from `predict.mlm'. Programmed and tested. * predict.q: Initial revision Wed Sep 9 11:26:54 1998 Berwin Turlach * First.lib.template: Changed to Release 0.8 and saved as snapshot. * vcov.q: Fixed but in tib.covmat (didn't take absolute values). Fixed typos in vcov.l1ce. Added code that calculates the degrees of freedom by calculating the trace of the "hat" matrix. The parameter s2hat was dropped, not the covaraince matrix is returned in a list as `cov.unscaled' together with the degrees of freedom (`df', vector with two values). Programmed and debugged vcov.l1celist. Tue Sep 8 10:47:42 1998 Berwin Turlach * First.lib.template: Changed to Release 0.7 and saved as snapshot. * vcov.q: Programmed and tested `vcov.l1ce'. Perhaps some thought is still needed for a sensible default for the parameter `s2hat'. * misc.q: Changed `[.l1celist' and `[[.l1celist'' to take into account the changes made to `l1ce' wrt parameters returned. * l1ce.q: Added check whether a dataframe is given down. Added check whether variables are left over after sweeping out. Changed what values are returned, these changes were necessary in order to be able to calculate the covariance matrix of the estimates. Mon Sep 7 07:18:08 1998 Berwin Turlach * First.lib.template: Changed to Release 0.6 and saved as snapshot. * l1ce.q: Changed X.to.C[,-name.matches] to X.to.C[,-name.matches,drop=F] so that the result remains a matrix. Explicitly added as.double() when the guess for the constrained parameter is given down to C. If a vector of constraints is given, then the routine uses now .C("mult_lasso") instead of repeated calls to .C("lasso") (saves initialisation work in the C routine). * deviance.q: Added `deviance.l1celist'. Sat Sep 5 06:49:59 1998 Berwin Turlach * First.lib.template: Changed to Release 0.5 and saved as snapshot. * residuals.q: Added `residuals.l1celist'. * print.q: Added `print.l1celist'. * plot.q: Changed to make use of the methods `aux' and `coef'. * misc.q: Changed `[[.l1celist' such that indexes out of bounds are correctly handled. Added `[.l1celist' and `aux' as generic method. `aux.l1celist' returns a matrix with the bounds (relative and absolute) and the Lagrangians. * l1ce.q: Renamed the parameter `constraint' to `bound'. * fitted.q: Added fitted.l1ce (practically fitted.default) and fitted.l1celist. * coef.q: Added coef.l1celist. * fitted.q: Initial revision Fri Sep 4 11:28:13 1998 Berwin Turlach * First.lib.template: Changed to Release 0.4 and saved as snapshot. * misc.q: First go at the function, adapted from [[.times. * print.q: Added constraint and Lagrangian to output. * misc.q: Initial revision Fri Sep 4 10:30:26 1998 Berwin Turlach * vcov.q: Preparation for implementing the vcov function. Checked in in order to rename all files from *.sf to *.q. * plot.q: First go at plot.l1celist. * l1ce.q: Added the possibility that constraint is a vector. In this case an object of class "l1celist" is returned. * vcov.q, plot.q: Initial revision Mon Aug 31 07:40:46 1998 Berwin Turlach * First.lib.template: Changed to Release 0.3 and saved as snapshot. * l1ce.sf: Renamed a variable from `m' to `p' so that it would not overwrite the variable holding the information about the model. * print.sf: Adapted from printed.lm, programmed and debugged. * print.sf: Initial revision * deviance.sf: Adapted from deviance.lm, programmed and debugged. * l1ce.sf: Renamed some components to have same names as in objects returned by lm. * coef.sf: Renamed beta to coefficients. * deviance.sf: Initial revision * l1ce.sf: Corrected an error in the way the formula were handled (noticable when weights were present). Added heaps of bells and whistles. * residuals.sf: Adapted from residuals.lm, programmed and debugged. Fri Aug 28 08:02:11 1998 Berwin Turlach * residuals.sf: Initial revision Sun Aug 16 02:50:33 1998 Berwin Turlach * First.lib.template: Changed to Release 0.2 and saved as snapshot. * coef.l1ce.sf: Renamed a parameter. * l1ce.sf: Corrected an error in the adjustment of the swept out parameters. Sun Aug 16 02:44:58 1998 Berwin Turlach * formula.util.sf: Corrected merge.formula such that it handles long formulae correctly. (deparse may return a *vector* of strings.) * coef.l1ce.sf: Initial revision Tue Aug 11 06:15:18 1998 Berwin Turlach * l1ce.sf: Complete rewrite to allow that several variables are swept out. * formula.util.sf: Some changes to merge.formula * formula.util.sf: Added merge.formula * formula.util.sf: Initial revision Sun Aug 9 02:57:17 1998 Berwin Turlach * l1ce.sf: A first version. Fri Aug 7 03:19:47 1998 Berwin Turlach * l1ce.sf: Initial revision Fri Aug 7 02:57:41 1998 Berwin Turlach * First.lib.template: Initial revision