The R Project SVN R-packages

Rev

Rev 1402 | Blame | Last modification | View Log | Download | RSS feed 

2005-05-19  Douglas Bates  <bates@bates1-home>

    * R/lmer.R (fn): Restored the ML in name MLdeviance from summary.

2005-04-18  Douglas Bates  <bates@bates1-home>

    * R/lmer.R: Choose method values appropriately for lmer with or
    without a family argument.

    * DESCRIPTION: change dependence to Matrix(>= 0.95-6)

    * R/lmList.R: Added argument pool to confint method for lmList and
    order to plot method for confint.lmList

2005-04-15  Douglas Bates  <bates@bates1-home>

    * DESCRIPTION: Change dependence to R (>= 2.1.0) and remove the
    contr.SAS definition and documentation.

2005-04-12  Douglas Bates  <bates@bates1-home>

    * R/lmer.R: Fixed problem with assignment of frame in lmer method
    with nonmissing family argument.

2005-04-06  Douglas Bates  <bates@bates1-home>

    * R/lmer.R: Fix with method to accomodate for nontrivial na.action

    * R/lmList.R: Repair the plot method for the lmList.confint class

2005-04-05  Douglas Bates  <bates@bates1-home>

    * R/lmer.R: Update of the "with" method.
    * : Removed humongous PDF file accidently included in tarball.
    Created release 0.95-3.
    

2005-02-27  Douglas Bates  <bates@bates1-home>

    * R/lme.R: Removed the stubs of the plot methods.

    * NAMESPACE: Remove imports from the Matrix package as it is
    listed in the Depends.

2005-02-24  Douglas Bates  <bates@bates1-home>

    * R/lmer.R (fn): Added an anova method for lmer objects that
    provides F tests based on the terms object.
    * data/ and man/: Moved all data sets from "Mixed-Effects Models
    in S and S-PLUS" to the MEMSS package and all other test data sets
    to the mlmRev package.

2005-01-24  Douglas Bates  <bates@bates7>

    * man/{guPrenat.Rd,guImmun.Rd,s3bbx.Rd,s3bby.Rd}: Add
    \encoding{latin1} per Kurt's request. 

2005-01-23  Douglas Bates  <bates@bates7>

    * R/lmer.R (fn): In the gradient function passed to optim, check
    if the parameters are equal to the current values before
    installing them.  This could save an unnecessary decomposition.

2005-01-22  Douglas Bates  <bates@bates7>

    * R/lmer.R (lmer): Changes in formulas code to allow explicit
    interactions in the grouping factors and formula without a fixed
    term.

2005-01-03  Douglas Bates  <bates@bates1-home>

    * R/lmer.R: Switch from lmeRep to lmer representation

2004-12-16  Douglas Bates  <bates@bates1-home>

    * R/lme.R (make.mf.call): Omit control values that are no longer used.

2004-12-13  Douglas Bates  <bates@bates1-home>

    * R/lme.R (lmeControl): Change msVerbose to an integer value.

    * R/ssclme.R: Change parameterization for optimization to
    constrained version that approaches the boundary safely.

    * tests/egsingle.R: Added regression test for multiple grouping
    factors with multivariate random effects.

2004-10-04  Douglas Bates  <bates@bates1-home>

    * R/lme.R: Corrected number of arguments to lmeRep_sigma

2004-10-02  Douglas Bates  <bates@bates1-home>

    * R/lme.R Re-order random when facs is reordered (Closes #10)
    Also add as.factor in creation of facs.  Previously using a
    grouping factor that was not a factor produced an obscure error
    message.

2004-09-27  Douglas Bates  <bates@bates1-home>

    * man/lme-class.Rd: Added documentation for methods with lmeRep in
    the signature.  Will need to remove all methods for ssclme objects
    when lmeRep can be used for general models.

2004-06-11  Douglas Bates  <bates@bates2_home>

    * inst/doc/MlmSoftRev.Rnw: Add new vignette

    * R/GLMM.R: protect verbose PQL output.

2004-06-08  Douglas Bates  <bates@stat.wisc.edu>

    * R/lme.R (lmeControl) R/GLMM.R: Change name to PQLmaxIt and
    change defaults for verbosity options to getOption("verbose")

2004-04-18  Douglas Bates  <bates@stat.wisc.edu>

    * R/ssclme.R: Fix coef<- so it really does use the unconstrained
    parameters.

2004-04-05  Douglas Bates  <bates@stat.wisc.edu>

    * NAMESPACE: Export the vcov methods.

2004-03-28  Douglas Bates  <bates@bates2_home>

    * R/lme.R: added a fitted method for lme.  Added dimnames and
    names to the value of the fixef and ranef methods for ssclme.

    * R/AllClass.R (lme): Added a fitted slot.

    * R/{lme.R,sparseGLMM.R}: Remove redundant call to ssclme_factor
    before ssclme_EMsteps

2004-03-18  Douglas Bates  <bates@stat.wisc.edu>

    * R/{pdMat.R,pdIdent.R,pdLogChol.R,pdMatrixLog.R,pdNatural.R},
      src/{pdMat.c,pdIdent.c,pdLogChol.c,pdNatural.c},
      tests/{pdCompSymm.R,pdDiag.R,pdIdent.R,pdLogChol.R,pdNatural.R},
      man/{pdMat-class.Rd,pdmatrix-class.Rd,corrmatrix-class.Rd,
      pdDiag-class.Rd,pdIdent-class.Rd,pdNatural-class.Rd,
      pdLogChol-class.Rd,coefGets.Rd,pdCompSymm-class.Rd,
      pdfactor-class.Rd,pdFactor.Rd,pdMatrix.Rd,
      pdBlocked-class.Rd},AllClass.R,AllGeneric.R:
    Moved the pdMat classes to the Matrix package.  Added a dependency
    on Matrix(>= 0.8-1)

2004-03-11  Douglas Bates  <bates@stat.wisc.edu>

    * NAMESPACE: export fixef and ranef (Erik Jørgensen)

2004-01-26  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION (Depends): Required that version 0.4-7 be used with
    R<=1.8.1.  Version 0.4-8 will require R>=1.9.0

2004-01-14  Douglas Bates  <bates@stat.wisc.edu>

    * src/glmmStruct.c (nlme_glmmLaplace_solveOnly): There should be a
    fix in here because dirtyStored is set to FALSE when part of the
    stored slot is dirty.

    * R/glmm.R: Undo fix for standard errors.  Doesn't work with
    method = 'Laplace'.

2004-01-13  Douglas Bates  <bates@stat.wisc.edu>

    * R/glmm.R: Correct the standard errors for glmm objects

2003-10-14  Douglas Bates  <bates@stat.wisc.edu>

    * R/reStruct.R: Missing a PACKAGE= in a .Call

2003-09-30  Douglas Bates  <bates@stat.wisc.edu>

    * src/reStruct.c (nlme_getFixDF): Commented out declarations of
    unused variables here and in other functions.

    * src/glmmStruct.c: Commented out nlme_glmm_calculateCorrection,
    which was not being used.

2003-09-25  Saikat DebRoy  <saikat@stat.wisc.edu>

    * man/LMEgradient.Rd: Add LMEhessian documentation.

    * INDEX: Update INDEX

    * R/pdDiag.R, R/pdLogChol.R, src/Makevars, src/pdLogChol.c, src/reStruct.c:
    Changes to make LMEhessian work for for one level models using pdDiag and method="ML"

2003-09-20  Saikat DebRoy  <saikat@stat.wisc.edu>

    * man/lmeLevel-class.Rd, man/reStruct-class.Rd, man/reStruct.Rd:
    Add documentation for analytic gradient

    * man/lmeControl.Rd: Add entry for analyticHessian argument.

    * DESCRIPTION: Update the date.

    * DESCRIPTION: Bump version.

    * ChangeLog: Update ChangeLog.

    * src/reStruct.c:
    Add a new function nlme_hessian_level that calculates the hessianArray
    slot for a particular level.

    Modify nlme_commonDecompose suitably to call nlme_hessian_level when
    analyticGradient is TRUE. If the method is ML, do an extra step of
    REML calculation for use in nlme_hessian_level.

    * R/pdDiag.R: Add an LMEhessian method for pdDiag.

    * R/reStruct.R: Add support for analytic hessian.

    * R/reLevel.R: Add an LMEhessian method.

    * R/lme.R: Add support for analyticHessian.

    * R/AllGeneric.R:
    Add LMEhessian generic. Add new argument analyticHessian to the
    reStruct generic.

    * R/AllClass.R:
    Add a hessianArray slot to lmeLevel and an analyticHessian slot to
    reStruct.

2003-09-01  Douglas Bates  <bates@stat.wisc.edu>

    * src/reStruct.c (nlme_reStructEMsteps): Correction to calculation
    of denominator df for a model with an intercept.

    * DESCRIPTION (Date): Version 0.3-7

    * R/reStruct.R: Use printCoefmat instead of print.coefmat

2003-08-27  Douglas Bates  <bates@stat.wisc.edu>

    * R/AAA.R: require methods package

2003-08-12  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION: Version 0.3-5

    * R/glmm.R, R/lme.R, R/reStruct.R: Add PACKAGE on all .Call

2003-08-11  Douglas Bates  <bates@stat.wisc.edu>

    * man/corMatrix.Rd: Add alias for method.
    
    * man/guImmun.Rd: docs for rural were missing

    * DESCRIPTION: Version 0.3-4

    * man/pdCompSymm-class.Rd, man/pdDiag-class.Rd,
    man/pdIdent-class.Rd, man/pdLogChol-class.Rd, man/pdMat-class.Rd,
    man/pdNatural-class.Rd, man/pdgradient.Rd, man/reStruct-class.Rd,
    man/reStruct.Rd, man/summary.reStruct-class.Rd, man/glmm-class.Rd,
    man/glmmStruct.Rd, man/groupedData-class.Rd, man/lmList-class.Rd,
    man/lmList.Rd, man/lme-class.Rd, man/lme.Rd,
    man/lmeLevel-class.Rd, man/logDet.Rd, man/logLik-methods.Rd,
    man/nlme-internal.Rd, man/nlmeInternal-class.Rd, man/GLMM.Rd,
    man/LMEgradient.Rd, man/LMEoptimizeGets.Rd, man/VarCorr-class.Rd,
    man/formulas.Rd: aliases for methods

2003-07-21  Douglas Bates  <bates@stat.wisc.edu>

    * R/glmm.R, R/reStruct.R: Complete the argument name change

    * DESCRIPTION: 0.3-3

    * R/AllGeneric.R, R/glmm.R, R/reStruct.R:
    consistent name for first argument of fixef and fixef<-

2003-07-21  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION (Version): 0.3-3

    * R/AllGeneric.R, R/glmm.R, R/reStruct.R:
    consistent name for first argument of fixef and fixef<-

2003-07-20  Douglas Bates  <bates@stat.wisc.edu>

    * man/GLMM.Rd:
    Simplify example with Laplace.  The ill-conditioned example takes too
    long.

    * man/fixed.effects.Rd: Added documentation for fixef<-.

2003-07-20  Douglas Bates  <bates@stat.wisc.edu>

    * man/GLMM.Rd:
    Simplify example with Laplace.  The ill-conditioned example takes too
    long.

    * man/fixed.effects.Rd: Added documentation for fixef<-.

2003-07-20  Douglas Bates  <bates@stat.wisc.edu>

    * man/fixed.effects.Rd: Added documentation for fixef<-.

2003-07-18  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION, R/AllGeneric.R, R/glmm.R, man/GLMM.Rd: release 0.3-2

2003-07-18  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/AllClass.R, R/reStruct.R:
    Remove the groups slot from the reStruct class. We do not
    currently use it and the information is also in the groups
    slot of relevel class.

2003-07-17  bates  <bates@stat.wisc.edu>

    * ChangeLog, DESCRIPTION: Release 0.3-1

    * man/VarCorr.Rd: Remove argument documentation from generic

2003-07-17  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/glmmStruct.c:
    Fix some problems in calculating second order Laplace
    approximation to the GLMM log-likelihood

    * src/reStruct.c, src/reStruct.h (nlme_reStruct_fitted, nlme_reStruct_fitted_internal):
    Allow specification of levels for which the fitted value is calculated

    * R/reStruct.R: Add a fixef<- method for reStruct

    * R/glmm.R: Add Laplace optimziation for GLMM

    * R/glmmStruct.R: Handle complicated response correctly

    * R/AllGeneric.R: Add a fixef<- generic

2003-07-15  Douglas Bates  <bates@stat.wisc.edu>

    * ChangeLog, DESCRIPTION, src/reStruct.c:
    Correct dynamic array allocation in getFixDF; release new version

    * src/reStruct.c (nlme_getFixDF):
    Allocate dynamic storage using Calloc and Free to
    conform to C90 specification.

2003-07-15  Douglas Bates  <bates@stat.wisc.edu>

    * src/reStruct.c (nlme_getFixDF): Allocate dynamic storage using
    Calloc and Free to conform to C90 specification.

2003-06-30  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION (Version): 0.2-2
    
    * R/reStruct.R, src/reStruct.c: Move most of getFixDF method for
    reStruct to C code.

2003-06-24  Saikat DebRoy  <saikat@stat.wisc.edu>

    * man/lme.Rd: Add documentation for model and x arguments.

2003-06-24  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION: 0.2-1 release

2003-06-24  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/glmmStruct.c:
    Add support for 2nd order log-likelihood calculation

2003-06-24  Douglas Bates  <bates@stat.wisc.edu>

    * R/zzz.R: Check if nlme is attached.

2003-06-24  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c:
    Make sure decomposed slot is not zero sized when doing predecomposition. Fix nlme_inner_perc_table

    * src/reStruct.h: Add nlme_decomposeChunk to the header file.

    * R/reStruct.R: Fix getFixDF. Set the assign.X slot.

    * R/lme.R:
    Make the final lme object creation a little more efficient. Use model and x arguments. Make nlm the default optimizer

    * R/glmmStruct.R: Add some auxiliary functions used by the C code

    * R/glmm.R: Try to save some copying by using nlme_replaceSlot

    * R/AllGeneric.R: Add model and x as arguments to lme

    * R/AllClass.R:
    Add a assign.X slot to reStruct (contains the assign attribute for fixed effects model.matrix.
    Add a fitted slot to lme.

2003-06-23  Douglas Bates  <bates@stat.wisc.edu>

    * src/utilities.h: declare nlme_check_Lapack_error

2003-06-17  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/lme.R, R/pdCompSymm.R, R/reStruct.R:
    Use better stepmax value for nlm.

    * R/reStruct.R: Fix problem when using numeric derivative.

    * src/pdLogChol.c: Fix problems for pdLogChol

    * src/pdDiag.c: Fix problems for pdDiag

2003-06-14  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION: New version to ship to Uwe Ligges

2003-06-12  Douglas Bates  <bates@stat.wisc.edu>

    * man/guImmun.Rd, tests/guImmun.R, INDEX, data/guImmun.rda, man/GLMM.Rd, man/glmm-class.Rd:
    New form of guImmun data

2003-06-12  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/pdIdent.R: Make sure the names propagate correctly for pdIdent.

    * R/VarCorr.R, R/reStruct.R:
    Ommit printing the correlation part when noCorrelation=TRUE for all levels.

    * R/reStruct.R, R/VarCorr.R:
    When noCorrelation=TRUE in summary.pdMat, do not print the correlations.

2003-06-12  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/glmm.R, R/lme.R: Fix typo.

    * INDEX: New version of INDEX

    * man/lmList-class.Rd, man/lmList.Rd, man/lme-class.Rd, man/nlmeInternal-class.Rd, man/VarCorr-class.Rd, man/VarCorr.Rd, man/family-class.Rd, man/glmm-class.Rd, man/glmmStruct.Rd, man/isInitialized.Rd:
    Add new documentation

2003-06-11  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/AllGeneric.R: Remove the response generic.

    * R/AllClass.R, R/AllGeneric.R, R/VarCorr.R:
    Add VarCorr class, generic and methods

    * R/pdNatural.R, src/pdNatural.c: Fix problems with pdNatural.

    * R/glmmStruct.R: Use getResponse instead of response.

    * R/glmm.R, R/lme.R:
    Do not show significance star inside show.lme or show.glmm.

2003-06-10  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/reStruct.R, R/lme.R: Remove extra right paren.

    * R/glmm.R:
    Change the display of family information in show.summary.glmm.

    * man/pdgradient.Rd, man/pdmatrix-class.Rd, man/solve-methods.Rd:
    Remove references to pdScalar

    * R/reStruct.R:
    Remove the response and groups methods. Add getResponse and getGroups methods.

    * R/pdDiag.R: Fix the check for length of nam in setAs

    * R/lme.R: Add coef and getGroups methods for the lme class

    * R/AllGeneric.R: Remove the groups generic

    * R/reStruct.R (show.summary.reStruct):
    Print the fixed effects formula.

    * R/glmmStruct.R:
    Only add second derivative information to family when using Laplace.
    Accept quasibinomial and quasipoisson families with second derivatives.

    * R/glmm.R (summary.glmm): Make useScale FALSE when appropriate.
    (show.summary.glmm): Print family and link information.

    * R/AllClass.R:
    Add a fixed slot to the reStruct class and method and family slots to the
    glmm class

    * src/reStruct.h: Add declaration for nlme_commonDecompose.

    * R/glmmStruct.R (glmmStruct): Add the method argument.

    * R/glmm.R:
    Add a new argument nEM.IRLS (the number of EM iterations in IRLS
    steps). Add summary method for glmm class and show methods for glmm
    and summary.glmm class.

    * R/AllGeneric.R:
    Change the signature of the glmmStruct generic to have a method
    argument.

    * R/AllClass.R:
    Add a summary.glmm class. Add a method slot to the glmm class.

    * R/reStruct.R (LMEoptimize):
    Return without doing anything if number of optimizer
    iterations is less than 1.

2003-06-06  Douglas Bates  <bates@stat.wisc.edu>

    * man/pdSymm-class.Rd, man/coefGets.Rd, man/isInitialized.Rd, man/lmeControl.Rd, man/logDet.Rd, man/pdLogChol-class.Rd, man/pdMat-class.Rd, man/pdNatural-class.Rd, R/glmm.R, man/EMupdateGets.Rd, man/GLMM.Rd, man/LMEgradient.Rd, INDEX, R/AllGeneric.R:
    release 0.1-19

2003-06-05  Douglas Bates  <bates@stat.wisc.edu>

    * man/GLMM.Rd: Rearrange argument sequence.

    * ChangeLog, DESCRIPTION: Release 0.1-19

    * R/glmmStruct.R: Explicit coercion for glmFit$y

    * R/lme.R, R/reStruct.R: Avoid copies of large object

    * R/pdMat.R, R/pdCompSymm.R, R/pdLogChol.R, R/pdNatural.R:
    summary method for pdMat class only

2003-06-05  Douglas Bates  <bates@stat.wisc.edu>

    * R/lme.R: Avoid copying the reStruct object when assigning the
    dontCopy slot.

    * R/pdMat.R: Remove defaults in summary method arguments.
    Build switch on noCorrelation into summary method for pdMat.
    Remove summary method for other classes that inherit from pdMat. 

2003-06-03  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION (Version): Versions 0.1-17 and 0.1-18 for debugging
    the methods package.
    
    * R/AllClass.R: remove the pdScalar class and definitions

2003-05-23  Douglas Bates  <bates@stat.wisc.edu>

    * src/utilities.h: Include the R_ext/Lapack.h file here; remove
    lapack.h and all references to it.

    * src/utilities.c (nlme_replaceSlot): Added C routine to replace the
    value of a slot in place.

2003-05-21  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/reStruct.R (reStruct):
    The reStruct class no longer has a frame slot.

    (response): Add a new extractor for the response.

    (fitted): Fitted no longer returns the fitted value in original order.

    * R/lme.R (lme):
    Rewritten to use the new definition of the lme class with
    reStruct as a slot.

    Add a coercion from lme to reStruct class (which just extracts the
    reStruct slot).

    (residuals): Rewrite to use the na.action slot if it is non NULL.

    (fitted): New method using the na.action slot if it is non NULL.

    (logLik): New method that calls the logLik method for the reStruct
    slot.

    * R/glmmStruct.R (glmmStruct):
    Now returns an object of class glmm which contains an
    reStruct class.

    * R/glmm.R (GLMM):
    Add a new argument called method with default value of
    "PQL". Rewrite the function to use this argument and also to reflect
    the fact that glmmStruct() now returns an object of class glmm with an
    reStruct slot.

    * R/AllGeneric.R (response): New generic added.

    * R/AllClass.R:
    Move the frame slot from the reStruct class to the lme class.
    Add an na.action slot to the lme class. It contains the na.action
    attribute of the model.frame and is used along with napredict in
    residuals and fitted methods.

    Rename the glmmStruct class to glmm. It now inherits from the lme
    class.

2003-05-21  Douglas Bates  <bates@stat.wisc.edu>

    * man/lmeScale.Rd: Now inlined in lmeControl

    * data/Machines.rda, data/MathAchSchool.rda, data/MathAchieve.rda, data/Oxboys.rda, man/bdf.Rd, data/bdf.rda:
    Already available in nlme package

    * man/lme.Rd, R/AllClass.R, R/reStruct.R:
    Summary and show methods for lme and reStruct

    * R/lme.R: Cleaned up the summary and show methods.

2003-05-21  Douglas Bates  <bates@stat.wisc.edu>

    * R/lme.R: Inlined the lmeScale function.  Modified the show
    method for the summary.lme and lme classes.

2003-05-21  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/pdNatural.R: Initialization should the same thing as in pdLogChol.

2003-05-19  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION, INDEX, data/guImmun.rda, man/EMupdateGets.Rd, man/glmmPQL.Rd, man/lmeScale.Rd, man/predecompose.Rd, man/reStruct-class.Rd, man/reStruct.Rd, tests/Bryk.R:
    release 0.1-15

    * R/AllGeneric.R: getFixDF made generic

    * R/pdDiag.R, R/pdIdent.R, R/pdLogChol.R, R/pdMat.R: Initialization

    * R/glmm.R, man/GLMM.Rd: rearranging arguments

    * R/AllClass.R, R/lmeFitted.R, R/lme.R, R/reStruct.R:
    New summary approach

    * tests/glmmPQL.R, tests/guImmun.R, tests/guPrenat.R, tests/Bryk.R, tests/GLMM.R, tests/bdf.R:
    *** empty log message ***

2003-05-19  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c: Use the dontCopy slot

    * src/glmmStruct.c: Bug fix

    * R/lmeFitted.R: Fix summary methods

2003-05-17  Douglas Bates  <bates@stat.wisc.edu>

    * tests/guPrenat.R, tests/guImmun.R: Adding tests

2003-05-16  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/lme.R:
    We were not adding the random effects terms to the formula for model.frame.

    * R/pdNatural.R, R/pdLogChol.R: Fix check for slot Ncol.

2003-05-14  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/glmm.R, R/glmmStruct.R, R/lme.R, src/STYLE:
    Remove code for 2nd order Laplace approximation for GLMMs from R functions

2003-05-14  Saikat DebRoy  <saikat@stat.wisc.edu>

    * Put a tag (laplace2-irls) for code with 2nd order Laplace
    computaion using IRLS steps for betas.

2003-04-16  Douglas Bates  <bates@stat.wisc.edu>

    * tests/MathAchieve.R: Test is redundant if Bryk test is used

    * man/lmeControl.Rd: analyticGradient argument

2003-04-15  Douglas Bates  <bates@stat.wisc.edu>

    * R/glmmStruct.R: Added mu2.eta2 for Poisson family with log link

2003-04-09  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c, R/reStruct.R, R/lmeFitted.R, R/glmm.R:
    Some bug fixes and enhancements

    * R/lmeFitted.R: There was a silly typo in the summary method for lme.

2003-04-07  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/lmeFitted.R:
    Add summary method (and supporting functions/methods) for lme objects

    * tests/Bryk.R, src/reStruct.h, src/reStruct.c, src/glmmStruct.c, R/reStruct.R, R/pdMat.R, R/pdCompSymm.R, R/lme.R, R/glmmStruct.R, R/glmm.R, R/AllGeneric.R, R/AllClass.R, ChangeLog:
    Large changes: check for NAMED tag whenever an reStruct gets modified.
    Also, modifications for correctly handling GLMMLaplace.

2003-03-26  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/glmmStruct.R: Use nlm by default for glmmLaplace.

2003-03-26  Douglas Bates  <bates@stat.wisc.edu>

    * INDEX, data/00Index: Version 0.1-13

    * R/pdMat.R, R/reStruct.R: Changes from print method to show method

    * R/lme.R: Added lme class

    * man/logDet.Rd, man/pdMat-class.Rd, man/pdSymm-class.Rd, man/pdgradient.Rd, man/pdmatrix-class.Rd, man/solve-methods.Rd, man/EMupdateGets.Rd, man/LMEgradient.Rd, man/coefGets.Rd, man/isInitialized.Rd:
    Removed references to pdMatrixLog class

    * man/Bryk.Rd: Added documentation for Bryk

2003-03-26  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/AllClass.R:
    Remove duplicate slot niterIRLS from definition of glmmStruct.

    * R/AllClass.R: Fix some conflicts with local modifications here.

2003-03-25  Douglas Bates  <bates@stat.wisc.edu>

    * R/AllClass.R:
    Removed EMiter slot from the reStruct class as it was not being used

    * man/pdMatrixLog-class.Rd: Move pdMatrixLog class to backup file

    * man/reStruct.Rd: incorporate the weights argument

    * R/pdMatrixLog.R: Move pdMatrixLog class to backup file

    * R/AllClass.R: Explicitly use contains in class definitions

    * install.R: methods package now attached by default

    * DESCRIPTION: New version

    * R/pdMatrixLog.bak:
    Moved pdMatrixLog class to a backup file.  We may want to remove this
    class.  There is no easy way to define the gradient for the MatrixLog
    parameterization of positive-definite symmetric matrices.

    * man/pdIdent-class.Rd:
    Changed pdDiag to pdIdent in places that had been copied from
    pdDiag-class.Rd

    * man/pdIdent-class.Rd: Filled in the skelton man page.

2003-03-18  Douglas Bates  <bates@stat.wisc.edu>

    * src/reStruct.c:
    Correction of commonDecompose for REML.  Value of columns slot was
    being used instead of its length.

2003-03-14  Douglas Bates  <bates@stat.wisc.edu>

    * src/reStruct.c: Calculation of constant

    * R/reStruct.R: Clean up EM iteration output

2003-03-13  Douglas Bates  <bates@stat.wisc.edu>

    * tests/bdf.R: Simplify call

    * R/reLevel.R: Introduced factor in calculation of initial value

2003-03-11  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.h, src/reStruct.c, src/glmmStruct.c, R/glmmStruct.R, R/AllGeneric.R, R/AllClass.R:
    Bug fixes. Do 2nd order laplace computations with numeric gradients (for now).

2003-03-10  Douglas Bates  <bates@stat.wisc.edu>

    * tests/Bryk.R, data/Bryk.rda: Pre DSC-2003

    * DESCRIPTION: New version before DSC-2003

2003-03-03  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.h, src/reStruct.c, src/glmmStruct.c, R/glmmStruct.R:
    Do the IRLS steps in a C function.

    * src/reStruct.h, src/reStruct.c, src/glmmStruct.c, R/reStruct.R, R/reLevel.R, R/glmmStruct.R, R/glmm.R, R/AllGeneric.R, R/AllClass.R:
    Fix bugs

2003-02-27  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION: New version

    * tests/MathAchieve.R: Use simpler form of random specification

2003-02-27  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/glmmStruct.R, R/glmm.R:
    Add a constructor function for glmmStruct class.

    * R/AllClass.R: Add two more slots to glmmStruct

2003-02-26  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.h, src/reStruct.c, src/glmmStruct.c:
    Add reStruct.h. Also make some bug fixes.

    * R/reStruct.R, R/glmmStruct.R:
    Some bug fixes. Also, add PACKAGE="lme4" to .Call calls.

    * src/reStruct.c: Add nlme_setParameters

    * src/glmmStruct.c: Add nlme_glmm_commonDecompose

    * R/glmmStruct.R: Some fixes

    * src/glmmStruct.c, R/glmmStruct.R:
    Add new files for handling glmmStruct

    * src/reStruct.c: Code restructure to handle GLMM 2-nd order Laplace

    * R/glmm.R: use glmmStruct as the class of result from glmmPQL

    * R/AllClass.R: Add new class glmmStruct

2003-02-13  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/reStruct.R: Do not return hessian from either optim or nlm calls.

2003-02-04  Douglas Bates  <bates@stat.wisc.edu>

    * src/lapack.h, DESCRIPTION, cleanup: cleanup no longer necessary

2003-01-30  Douglas Bates  <bates@stat.wisc.edu>

    * src/reStruct.c: Fixed up logic in nlme_logPenalty.

    * cleanup: No longer want to remove src/Makevars

    * configure.ac, configure.win, data/00Index, configure, aclocal.m4, src/Makevars, src/Makevars.in, src/Makevars.win:
    Switching to $(LAPACK_LIB) for R-1.7.0

2003-01-29  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION (Date): Create a new version

    * cleanup: Changed to using the $(LAPACK_LIB) macro from R-1.7.0.
    Removed the files to generate the configure script.

2003-01-12  Saikat DebRoy  <saikat@stat.wisc.edu>

    * data/00Index, INDEX: Commit result from R CMD build --force

    * ChangeLog: Cleanup ChangeLog file.

2003-01-02  Douglas Bates  <bates@stat.wisc.edu>

    * tests/glmmPQL.R: added the standard deviations and their
    logarithms to the results.

    * data/guImmun.rda, data/guPrenat.rda: Added the Immunization and
    Prenatal care data sets from Rodriguez and Goldman (2001).

2002-12-29  Douglas Bates  <bates@stat.wisc.edu>

    * INDEX: Automatic INDEX update

    * DESCRIPTION: lme4 now requires only R-1.6.2

    * src/utilities.h and many otherss:
    Move dtrtri and dtrmm declarations to .h file.  Change all
    check_Lapack_error calls to nlme_check_Lapack_error. Remove
    methods.h and all references to it.

    * man/BIC.Rd, man/QR-class.Rd, man/summary.reStruct-class.Rd:
    More documentation modifications

    * R/reStruct.R: Create a logLik object from the logLik method

    * R/AllGeneric.R: Add the BIC generic

    * R/AllClass.R, src/qr.h, src/qr.c: Remove the QR class

2002-12-29  Douglas Bates  <bates@stat.wisc.edu>

    * man/BIC.Rd, man/QR-class.Rd, man/summary.reStruct-class.Rd:
    More documentation modifications

    * R/reStruct.R: Create a logLik object in the logLik method

    * R/AllGeneric.R: Add the BIC generic

    * R/AllClass.R: Remove the QR class

    * data/s3bbx.rda, data/s3bby.rda: compress these large data objects

2002-12-24  Douglas Bates  <bates@stat.wisc.edu>

    * tests/glmmPQL.R: Change test to provide timings.

    * tests/pdNatural.R: Updates of pdNatural class

    * INDEX: automatic update by R CMD build

    * R/pdNatural.R, src/pdNatural.c: Updates of pdNatural class

2002-12-18  Douglas Bates  <bates@stat.wisc.edu>

    * DESCRIPTION: Increment the version, change the description and title

2002-12-18  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c:
    Change nlme_logLikelihhod and nlme_commonDecompose so that they accept an
    optional parameter vector. Also use the dirtyStored flag to avoid doing
    redundant computations.

    * R/reStruct.R:
    Modifications for adding the dirtyStored slot of reStruct objects.

    Set parameter values where possible by calls to C functions
    nlme_logLikelihood and nlme_commonDecompose. Alternative apporoach of using
    coef<- calls may produce extra copies of reStruct objects.

    * R/AllClass.R:
    Add a new slot (dirtyStored) to the reStruct object with its default value
    in the prototype being FALSE.

2002-12-05  Douglas Bates  <bates@stat.wisc.edu>

    * R/reStruct.R: Added a rudimentary summary method.

2002-12-03  Douglas Bates  <bates@stat.wisc.edu>

    * R/pdDiag.R, R/pdLogChol.R, R/pdScalar.R:
    package in .Call should be PACKAGE

2002-12-01  Douglas Bates  <bates@stat.wisc.edu>

    * man/logLik-methods.Rd, man/matrixGets.Rd, man/pdgradient.Rd, man/pdmatrix-class.Rd, man/lmeScale.Rd:
    Cleaning up manual pages

2002-11-30  Douglas Bates  <bates@stat.wisc.edu>

    * R/pdScalar.R: Corrected the calculation for the EM update.

2002-11-30  saikat  <saikat@stat.wisc.edu>

    * ChangeLog: Clean ChangeLog.

2002-11-30  Douglas Bates  <bates@stat.wisc.edu>

    * R/pdScalar.R: added coersion of value in EMupdate<-

    * tests/glmmPQL.R: *** empty log message ***

    * man/model.matrixGets.Rd, man/responseGets.Rd:
    Added more documentation of generics

    * man/reStruct-class.Rd: Remove redundant \alias{reStruct}

    * tests/glmmPQL.R: Change number of fits to 1.

    * man/pdDiag-class.Rd: Update method descriptions

    * INDEX: *** empty log message ***

    * man/lmeScale.Rd: Transferred from nlme package

    * R/pdMat.R: Fix problem with logDet slot of pdfactor

    * R/pdScalar.R: Filter all changes through coef<-

    * R/pdDiag.R: Fix and error message and remove redundant checks

2002-11-30  Douglas Bates  <bates@stat.wisc.edu>

    * man/reStruct-class.Rd: Remove redundant \alias{reStruct}

    * tests/glmmPQL.R: Change number of fits to 1.

    * man/pdDiag-class.Rd: Update method descriptions

    * INDEX: *** empty log message ***

    * man/lmeScale.Rd: Transferred from nlme package

    * R/pdMat.R: Fix problem with logDet slot of pdfactor

    * R/pdScalar.R: Filter all changes through coef<-

    * R/pdDiag.R: Fix an error message and remove redundant checks

2002-11-29  Douglas Bates  <bates@stat.wisc.edu>

    * tests/pdScalar.R: added pdScalar methods

    * src/pdDiag.c, src/pdLogChol.c:
    Ensure duplicate if creating new object for slot

    * src/pdScalar.c: added pdScalar methods

    * tests/glmmPQL.R: default number of glmmPQL tests is 5

    * man/lme.Rd, man/lmeLevel.Rd, man/matrixGets.Rd, man/pdDiag-class.Rd, man/getCovariate.Rd, man/getGroups.Rd, man/getGroupsFormula.Rd, man/glmmPQL.Rd, man/groups.Rd, man/isInitialized.Rd, man/MathAchieve.Rd, man/Names.Rd, man/coefGets.Rd, man/corFactor.Rd, man/corMatrix.Rd, man/fixed.effects.Rd, man/EMstepsGets.Rd, man/EMupdateGets.Rd, man/LMEgradient.Rd, man/LMEoptimizeGets.Rd:
    \docType{genericFunction} for generic/method docs

    * src/pdDiag.c, R/pdDiag.R: coef<- method as .Call

    * R/pdLogChol.R: Documentation clarification

2002-11-28  Douglas Bates  <bates@stat.wisc.edu>

    * R/reLevel.R: Use TRUE and FALSE instead of T and F.

2002-11-28  saikat  <saikat@stat.wisc.edu>

    * cleanup: Remove *.o and *.so files in source during cleanup.

    * ChangeLog: Remove duplicate entries.

2002-11-27  Douglas Bates  <bates@stat.wisc.edu>

    * data/00Index: Updates for the Rodriguez/Goldman data

    * ChangeLog: *** empty log message ***

    * INDEX: Update from new man pages

    * DESCRIPTION: Minor release increment

    * data/s3bby.rda, data/s3bbx.rda:
    Added Rodriquez/Goldman simulated data

    * tests/glmmPQL.R:
    Added glmmPQL test for Rodriquez/Goldman simulated data

    * man/random.effects.Rd, man/reStruct.Rd, man/s3bbx.Rd, man/s3bby.Rd, man/glmmPQL.Rd, man/groupedData-class.Rd, man/lme.Rd, man/logLik-methods.Rd, man/nfGroupedData-class.Rd, man/nffGroupedData-class.Rd, man/nfnGroupedData-class.Rd, man/nmGroupedData-class.Rd, man/EMstepsGets.Rd, man/LMEoptimizeGets.Rd, man/QR-class.Rd, man/fixed.effects.Rd:
    Updated manual pages

    * R/reStruct.R: Added fixef and ranef methods.

    * R/AllGeneric.R:
    Added generics for fixed.effects, fixef, random.effects and ranef.

    * ChangeLog: *** empty log message ***

    * man/pdLogChol-class.Rd, man/pdfactor-class.Rd, man/pdmatrix-class.Rd, man/reStruct-class.Rd, man/reStruct.Rd, man/solve-methods.Rd, man/weightedGets.Rd, src/reStruct.c, INDEX, man/lmeControl.Rd, man/lmeLevel-class.Rd, man/lmeLevel.Rd, man/logDet.Rd:
    Update help pages

    * data/00Index, data/dallas.rda:
    Store ids and Campus as factors in dallas data set

    * man/reStruct-methods.Rd: Method descriptions moved to reStruct.Rd

    * man/logDet-methods.Rd: Method descriptions moved to logDet.Rd

    * man/weightedGets-methods.Rd:
    Method descriptions moved to weigghtedGets.Rd

    * man/lmeLevel-methods.Rd: Method descriptions moved to lmeLevel.Rd

2002-11-27  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c (nlme_predecompose):
    Look at the useWeighted slot to decide if the
    weighted or original slot should be decomposed.

    * R/AllClass.R:
    Add a new slot, useWeighted to reStruct class. This specifies whether
    the weighted slot should be used for computations.

    * R/reStruct.R:
    (reStruct), (response<-), (model,matrix<-): Deal with the new
    useWeighted slot of reStrict.

    (weighted<-): Now value is a numeric vector of weights. Try to update
    the weighted slot in place if we can. If all the weights are one, set
    useWeighted to FALSE.

    * R/lme.R (lmeControl):
    Add new option EMverbose for a verbose mode for EM
    steps.

    * R/glmm.R (glmmPQL):
    Raise default value of niter to 20. Also change default
    value of verbose to FALSE.

    Take only a single EM step (ie. control$niterEM=1) after the first
    iteration. Taking too many EM steps after the first iterations seems
    to be unnecessary.

    Make the verbose flag also print loglikelihood value and parameter
    values in the iteration step.

    * src/reStruct.c (nlme_reStruct_fitted):
    Do not use weighted for computing fitted values.

2002-11-26  Douglas Bates  <bates@stat.wisc.edu>

    * man/pdFactor.Rd, man/pdMat-class.Rd, man/pdMatrix-methods.Rd, man/pdMatrix.Rd, man/predecompose-methods.Rd, man/predecompose.Rd, man/matrixGets.Rd, man/pdFactor-methods.Rd, man/coefGets.Rd, man/corMatrix.Rd:
    Cleaning man pages

2002-11-26  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c:
    Use Calloc/Free instead of using allocMatrix/allocVector for all
    temporary objects.

    Update doxygen comments.

    * src/bitfield.c (nlme_bitfield_alloc):
    No need to set to zero when memory is allocated
    with Calloc.

2002-11-25  Douglas Bates  <bates@stat.wisc.edu>

    * man/dallas.Rd, man/groups-methods.Rd, man/groups.Rd, man/isInitialized-methods.Rd, man/isInitialized.Rd, man/EMupdate.Rd, man/EMupdateGets-methods.Rd, man/EMupdateGets.Rd, man/LMEgradient-methods.Rd, man/LMEgradient.Rd, man/Names.Rd, man/coefGets-methods.Rd, man/coefGets.Rd, man/corFactor.Rd:
    consolidate generic and methods documentation when there are no surprising arguments

2002-11-25  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/glmm.R:
    First cut at a glmmPQL function mostly borrowed from MASS and adapted
    to new setup. I do not think it is working correctly right now -
    results with the rg3 data do not match fit obtained from the version
    in MASS.

    * src/reStruct.c (nlme_decomposeChunk):
    Be more careful about calls to copyRows - under
    certain conditions we were trying to copy more than what was actually
    in the chunk.

    * R/reStruct.R (reStruct): Initialize the new origOrder slot.

    (coef<-): Set the names in value to NULL - otherwise names were
    becoming longer and longer under certain situations.

    (response<-): Correct indexing while setting the response in original
    slot.

    * R/AllClass.R:
    Add a new slot origOrder to reStruct class. It stores the ordering
    used to go from original order to the one used internally in reStruct.

2002-11-24  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/qr.h: Include methods.h for SEXPREC typedef

    * src/methods.h: const SEXP is not meaningful - just use SEXP.

    * src/lapack.h (dgemv): y should not be const.

    * src/pdNatural.c, src/pdMat.c, src/pdLogChol.c, src/pdIdent.c, src/pdDiag.c:
    Some code cleanup.

    * src/bitfield.c:
    (nlme_bitfield_alloc), (nlme_bitfield_free): Use Calloc/Free.

    * src/reStruct.c: Add some documentation.

    Use direct call to LAPACK QR routines. Use Calloc/Free for most
    temporary memory allocation. The only big one still allocated via
    allocVector is the scratch array - mostly because of the way copyRows
    and zeroRows are written.

    (nlme_reStruct_fitted): Add new function for finding the fitted values
    for reStruct objects.

    * R/reStruct.R (reStruct): The reverseOrder slot was wrong earlier.

    (reStruct), (model.matrix<-): Set the weighted slot to be a zero
    dimension matrix unless we really need it.

    (response<-): Add new method for setting the response.

    (fitted): Add new method for reStruct objects.

    * R/AllGeneric.R (response<-): Added new generic.

2002-11-23  Douglas Bates  <bates@stat.wisc.edu>

    * data/dallas.rda: TAAS data from Dallas Indep. School District

    * R/pdLogChol.R: Replaced the R code for coef<- with a .Call

    * src/pdLogChol.c:
    Remove pdLogChol_pdfactor and insert pdLogChol_coefGets using a static
    double function ld_factor_from_par.

    * src/reStruct.c (nlme_logLikelihood):
    Use the factor and logLik slots of the precision matrices instead of
    AS(precision, pdfactor_class).  Speeds up the calculation somewhat.

    * R/pdMat.R, src/pdIdent.c, src/pdLogChol.c, R/AllClass.R,
    R/pdDiag.R, R/pdIdent.R, R/pdLogChol.R: Added slots to pdMat class
    - removed them from inheriting classes 

    * tests/pdCompSymm.R, tests/pdIdent.R, tests/pdNatural.R:
    Adjustments for changes in slots.
    
    * tests/MathAchieve.R, tests/MathAchive.R: Corrected spelling of
    file name.

2002-11-22  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c (nlme_factor_level):
    Remove extra declaration of srcStartRow.

    * src/reStruct.c (nlme_factor_level):
    Fix bug when dealing with more than one levels.

    Remove some commented out parts.

    * R/reStruct.R (reStruct): Modify self@groups for multiple levels.

    * R/reLevel.R (lmeLevel):
    Fix problem with creating 1x1 matrices using diag.

2002-11-21  Douglas Bates  <bates@stat.wisc.edu>

    * INDEX, src/lapack.h, src/reStruct.c, DESCRIPTION:
    moving declarations in reStruct.c to lapack.h; bump the version number

2002-11-20  Douglas Bates  <bates@stat.wisc.edu>

    * man/solve-methods.Rd, man/weightedGets-methods.Rd, man/coefGets-methods.Rd, man/groups-methods.Rd, man/isInitialized-methods.Rd, man/lmeLevel-methods.Rd, man/logDet-methods.Rd, man/logLik-methods.Rd, man/pdFactor-methods.Rd, man/pdMatrix-methods.Rd, man/predecompose-methods.Rd, man/reStruct-methods.Rd, man/EMupdateGets-methods.Rd, man/LMEgradient-methods.Rd:
    Added more manual pages

    * man/pdScalar-class.Rd, man/pdgradNumeric.Rd, man/pdgradient.Rd, man/reStruct-class.Rd, man/weightedGets.Rd, man/QR-class.Rd, man/corMatrix.Rd, man/lmeLevel-class.Rd, man/logDet.Rd, man/matrixGets.Rd, man/pdBlocked-class.Rd:
    Cleaned up manual pages

2002-11-20  Douglas Bates  <bates@stat.wisc.edu>

    * man/pdgradient.Rd: Cleaned up this and several other help pages.

2002-11-20  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/qr.c (nlme_QR): Was unprotecting too many things.

    * src/pdLogChol.c (pdLogChol_pdfactor):
    Avoid making two copies of factor - SET_SLOT
    duplicates it in any case.

    * src/bitfield.c: Was including the main() definition by mistake.

    * src/qr.c (nlme_QR):
    Do not compute qr_class everytime - instead make it static
    and compute in the first call. This reduces run time considerably -
    MAKE_CLASS has a large overhead.

    SET_SLOT() duplicates its value. So avoid making two copies of mat and
    rcond as these can be large.

    * src/reStruct.c:
    Reuse scratch array for all chunks in a level inside nlme_predecompose
    and nlme_logLikelihood.

    (nlme_scratchSize): Calculate size of the scratch array.

2002-11-20  Saikat DebRoy  <saikat@stat.wisc.edu>

    * lme.R (lmeControl): Add another argument, optimizer, to the control
    list. Value should be either "nlm" or "optim".

    * reStruct.R (LMEoptimize<-):
    Allow use of either nlm or optim for optimization.

2002-11-19  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c (nlme_reStructDims):
    Fix bug with different lengths in chunks.

2002-11-19  Douglas Bates  <bates@stat.wisc.edu>

    * data/bdf.rda, man/MathAchSchool.Rd, man/MathAchieve.Rd, man/bdf.Rd, ChangeLog, DESCRIPTION, INDEX, R/lme.R, data/00Index, data/MathAchSchool.rda, data/MathAchieve.rda:
    Added data sets; minor fix in reStruct

    * R/reStruct.R:
    Minor fix in LMEoptimize - optimization result had wrong name.

    * R/lme.R (lmeScale): drop unused levels in call to model frame.
    Use the integer niterEM instead of the list in the replacement
    EMsteps. 

2002-11-19  Saikat DebRoy  <saikat@stat.wisc.edu>

    * lme.R (lme):
    Rename EMniter<- to EMsteps<- and replace call to optim by
    LMEoptimize<-.

    * reStruct.R (EMsteps<-): Renamed from EMniter<- and rewritten.

    (LMEoptimize<-): New - encapsulates nonlinear optimization step.

    * AllGeneric.R (EMsteps<-): Renamed from EMniter<- and rewritten.

    (LMEoptimize<-): New - encapsulates nonlinear optimization step.

    * reStruct.c (nlme_predecompose):
    Was decomposing original - it should be weighted.

2002-11-18  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reLevel.R (LMEgradient):
    Add new method for calculating gradient - uses the
    updateFactor slot instead of using the A argument.

    * reStruct.R (EMniter<-):
    Add new method for doing certain number of EM iterations
    in the reStruct object.

    (LMEgradient): Add new method for calculating gradient of an reStruct
    object.

    * pdScalar.R, pdNatural.R, pdMatrixLog.R, pdMat.R, pdLogChol.R, pdDiag.R, pdCompSymm.R:
    Replace as(x, foo) where foo is basic types like "numeric" or
    "logical" to as.foo(x) - for some reason the as() calls are too
    expensive now.

    * lme.R (lmeScale): Added from nlme-3.1

    (lmeControl): Added from nlme-3.1

    (lme): Add preliminary versions - right now just returns the fitted
    reStruct object.

    * AllClass.R: Add class definition for groupedData objects.

    * reStruct.R (coef<-):
    Should only set parameter values for random effects

    * reStruct.c (nlme_logLikelihood):
    Add missing PROTECT() around calculation of Delta.

2002-11-17  Douglas Bates  <bates@stat.wisc.edu>

    * pdLogChol.R: fix coef<-

2002-11-15  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.R (reStruct): Initialize storage for the bbetas slot.

    (predecompose): nlme_predecompose now checks the dirty slot - no need
    to check it here.

    (logLik): No need to call predecompose on reStruct object - done
    inside C routine.

    * reLevel.R (lmeLevel): Add support for nrow and updateFactor slots in
    constructor.

    (EMupdate<-): Must return x after modification.

    * AllClass.R: Add a new slot updateFactor to lmeLevel objects.

2002-11-14  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.R:
    do not modify stored when only calculating log-likelihood from R

2002-11-14  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.R (coef): remove the unconstrained parameter.

    (coef): Use new structure of reStruct objects with columns slot being
    inside the lmeLevel objects.

    (coef<-): Use new structure of reStruct objects with columns slot being
    inside the lmeLevel objects.

    * reLevel.R (coef): Added new method for class lmeLevel.

    (coef<-): Added new method for class lmeLevel.

2002-11-12  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.R (reStruct):
    changes for fixing problem with stored and decomposed
    indexes.

    * reLevel.R (lmelevel):
    Add a new method which has precision and groups arguments
    missing. This is used for fixed effects and response levels.

2002-11-11  Douglas Bates  <bates@stat.wisc.edu>

    * pdMat.R:
    print method for pdMat is causing codoc to fail - dont know why

    * pdLogChol.R: pdLogChol pdgradient and LMEgradient cleanup

    * pdDiag.R: pdDiag is failing R CMD check now

2002-11-09  Douglas Bates  <bates@stat.wisc.edu>

    * pdLogChol.R: Added PACKAGE= to .Call

    * pdLogChol.R: More pdLogChol modifications

2002-11-07  Douglas Bates  <bates@stat.wisc.edu>

    * AllClass.R, pdLogChol.R, pdMat.R: testing pdLogChol

2002-11-07  Douglas Bates  <bates@stat.wisc.edu>

    * src/pdLogChol.c (pdLogChol_pdgradient): added function.

2002-11-04  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/reStruct.R, R/pdNatural.R, R/pdMat.R, R/pdLogChol.R, R/pdIdent.R, R/pdDiag.R, R/pdCompSymm.R:
    Add PACKAGE="lme4" to .Call calls

    * R/zzz.R:
    Remove resetGeneric call - bug fixed in R 1.7.0 development version.

    * src/reStruct.c:
    Most non-static functions now have an reStruct object as argument and
    not individual slots.

    * src/qr.c, src/pdNatural.c, src/pdLogChol.c, src/pdIdent.c, src/pdDiag.c:
    Fix some compile warnings

    * R/AllClass.R: Add the bbetas slot to reStruct.

    * R/AllGeneric.R: Add a new predecompose generic.

    * R/reStruct.R: Add predecompose and logLik methods.

2002-10-31  Douglas Bates  <bates@stat.wisc.edu>

    * src/pdNatural.c, tests/pdLogChol.R, tests/pdNatural.R, R/pdNatural.R, src/pdLogChol.c:
    Adding pdNatural, tweaking pdLogChol

2002-10-30  Douglas Bates  <bates@stat.wisc.edu>

    * R/pdDiag.R, R/pdLogChol.R: pdLogChol fixes

    * src/Makevars: remove src/Makevars as it is automatically created

    * R/pdLogChol.R, R/pdMat.R:
    pdLogChol added (without LMEgradient and EMupdate)

2002-10-29  Douglas Bates  <bates@stat.wisc.edu>

    * src/pdIdent.c, src/methods.h, src/pdDiag.c:
    gradient methods added for pdDiag and pdIdent

2002-10-24  Douglas Bates  <bates@stat.wisc.edu>

    * R/pdMat.R, src/pdDiag.c, src/pdIdent.c, tests/pdIdent.R, R/AllClass.R, R/AllGeneric.R, R/pdDiag.R, R/pdIdent.R:
    Adding methods to the pdIdent class

    * R/reLevel.R, R/zzz.R, src/pdDiag.c, tests/pdDiag.R, R/AllGeneric.R, R/pdDiag.R, R/pdLogChol.R, R/pdMat.R, R/pdNatural.R, cleanup, configure.ac, ChangeLog, aclocal.m4:
    Cleaning up pd classes

2002-10-23  Douglas Bates  <bates@stat.wisc.edu>

    * configure.ac: Changed the configure.ac and aclocal.m4 files
    to the new versions contributed by Kurt Hornik for the Matrix
    package.

2002-10-22  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.R, reLevel.R, AllClass.R: rename reLevel to lmeLevel

2002-10-22  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.c (nlme_estimate_level):
    Other than possible bugs, this should be a
    working version.

    Fix compilation errors. Not all C functions do the correct things
    right now - but they do compile.

2002-10-18  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.c: More reLevel reorganization.

    * reStruct.c:
    Move columns, originalRows, decomposedRows and storedRows inside
    reLevel class.

2002-10-17  Douglas Bates  <bates@stat.wisc.edu>

    * qr.c, reStruct.c: Minor compilation bugs

2002-10-15  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.c: Add some skeleton code for EM update.

    * methods.h: Remove pre-1.6.0 definitions for GET_SLOT and SET_SLOT.
    .

    * bitfield.c (nlme_word_log2):
    Keep the old array based bit index lookup routine in
    case we need it later.

    * qr.c: change getAttrib(foo, R_DimSymbol) to GET_DIM(foo).

2002-10-15  Douglas Bates  <bates@stat.wisc.edu>

    * qr.c: Modifications to get past R CMD check

2002-10-15  Douglas Bates  <bates@stat.wisc.edu>

    * R/pdDiag.R: Spelling corrections for 'Uninitialized'.  Other
    changes to get a clean load into R.

2002-10-01  Saikat DebRoy  <saikat@stat.wisc.edu>

    * pdMat.R, pdLogChol.R, pdIdent.R, pdDiag.R, AllGeneric.R, AllClass.R:
    Cleaning up of pdMat class codes

2002-09-23  Saikat DebRoy  <saikat@stat.wisc.edu>

    * AllClass.R:
    No need for declaring old classes like data.frame as that is done in
    methods now.

2002-09-21  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c, src/lapack.h, src/bitfield.c:
    changes for compatibility with Doxygen

    * Doxyfile: Added for generation of Doxygen documentation

2002-09-20  Douglas Bates  <bates@stat.wisc.edu>

    * R/AllClass.R: comment style changes

2002-09-16  Saikat DebRoy  <saikat@stat.wisc.edu>

    * zzz.R:
    No need to do a library(methods) or require(methods) from inside
    .First.lib as we do that in install.R now.

2002-09-16  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.c (nlme_validateIndex): was called nlme_validIndex
    (nlme_validateRowIndex): validateIndex should be called nlme_validateIndex
    (nlme_validateColIndex): validateIndex should be called nlme_validateIndex

2002-09-13  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.c:
    Remove _range from name of set and unset bitfield functions.

2002-09-06  Saikat DebRoy  <saikat@stat.wisc.edu>

    * bitfield.h, bitfield.c:
    Rename nlme_bitfield_set_range, nlme_bitfield_unset_range and
    nlme_bitfield_toggle_range by dropping _range at end.

    * bitfield.c (nlme_bitfield_set): Use memset instead of a for loop
    * bitfield.c (nlme_bitfield_unset): Use memset instead of a for loop

2002-09-03  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.c: Fixed some bugs.

    * reStruct.c:
    Added preliminary version of nlme_reStructDims. Currently it only
    initializes originalRows slot.

2002-09-03  Saikat DebRoy  <saikat@stat.wisc.edu>

    * reStruct.c: Fixed some bugs.

    * reStruct.c:
    Added preliminary version of nlme_reStructDims. Currently it only
    initializes originalRows slot.

2002-09-03  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/AllClass.R:
    Use setOldClass to define data.frame, formula, factor and ordered.

    * src/reStruct.c:
    nlme_decomposeChunk: Only store first q rows after decomposition when
    storemat != NULL.

2002-09-02  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c: Value of ncolRest was wrong.

    * src/reStruct.c:
    Support appending Delta below Z in nlme_decomposeChunk().

    * src/qr.c:
    Make nlme_qy() add zero rows below qin if its number of rows is not
    big enough.

    * src/reStruct.c:
    Take out code in innermost loop of nlme_predecompose() and create new
    function nlme_decomposeChunk(). Modify the code so that it has stored
    and Delta in the signature (these are ignored right now).

    * src/bitfield.c:
    Was referring to range in nlme_bitfield_next_range() as if it was a
    structure and not a pointer.

2002-08-29  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/reStruct.R: Added missing @random in coef method for reStruct.

    * src/bitfield.c: Fix spelling mistake in comment

    * src/bitfield.c:
    Check for NULL range and x in nlme_bitfield_next_range()

    * src/reStruct.c:
    Was not checking if destNonZero was non NULL in nlme_copyRows()

    * src/reStruct.c, src/qr.h, src/qr.c:
    change "const SEXP" to "const SEXPREC*" : const of pointer typedefs are not
     pointer to constants

    * src/reStruct.c, src/qr.h, src/qr.c, src/bitfield.h, src/methods.h, src/bitfield.c:
    Some bugfixes

    * src/reStruct.c: use bitfields to keep track of zero rows

    * src/bitfield.c, src/bitfield.h: Bitfield handling rotines

2002-08-23  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/reStruct.c:
    Add nlme_predecompose() and routines needed by it. Some of those
    routines may move to a different place later.

    The function definitions are still incomplete.

    * src/methods.h: Did not have a declaration for nlme_new()

    * src/qr.c: Include qr.h in qr.c

    * src/qr.h: Create header file QR routines.

    * R/AllClass.R: Add reverseOrder slot to reStruct.

    * R/reStruct.R:
    Fixed some problems in creating row indices in self@dims.

2002-08-21  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/reLevel.R: Add reLevel generic.

    Remove coef method for reLevel - not required.

    * R/AllClass.R: Add definition of reLevel class.

    * R/reStruct.R:
    Use reLevel objects in the random slot of reStruct objects.

    * R/reLevel.R: Checked in first version of reLevel methods.

    * src/qr.c: Fixed nlme_new - now evals a call to new.

2002-08-14  Douglas Bates <bates@stat.wisc.edu>

    * src/pdMat.c, src/qr.c, R/AllClass.R, src/lapack.h:
    Modified slot names in QR class, added nlme_qy to qr.c

    * R/pdDiag.R, R/pdLogChol.R: Updates on pdDiag and pdLogChol

2002-08-12  Douglas Bates <bates@stat.wisc.edu>

    * R/pdNatural.R, src/pdDiag.c, src/pdMat.c, tests/pdDiag.R, R/AllGeneric.R, R/pdBlocked.R, R/pdCompSymm.R, R/pdDiag.R, R/pdIdent.R, R/pdMat.R:
    split R sources, added test methods for LMEgradient and EMupdate

2002-08-01  Saikat DebRoy  <saikat@stat.wisc.edu>

    * src/qr.c: added PROTECT to creation of qr

    * src/qr.c: compute receprocal of the condition number in nlme_QR

    * src/methods.h, src/qr.c: make NEW set the class

    * src/qr.c: add missing semicolon

    * R/AllClass.R, src/qr.c:
    rename invcond slot to rcond. compute rcond using dtrcon

    * src/qr.c: add an invcond slot to QR

    * R/zzz.R:
    do library(methods) from .First.lib so that "R INSTALL -s pdMat" works

    * INDEX, R/AllGeneric.R:
    do not use return types for generics - causes error

    * R/pdMat.R, R/reStruct.R, src/methods.h, src/qr.c: Fix problems in QR

2002-07-31  Saikat DebRoy  <saikat@stat.wisc.edu>

    * R/AllClass.R, R/AllGeneric.R, src/methods.h, src/pdMat.c, src/qr.c:
    Added a QR class

2002-07-30  Douglas Bates <bates@stat.wisc.edu>

    * R/AllClass.R, R/AllGeneric.R, R/pdMat.R, R/reStruct.R:
    Modifications from Saikat

    * DESCRIPTION, INDEX, R/AllClass.R, R/AllGeneric.R, R/pdMat.R, R/zzz.R, aclocal.m4, cleanup, configure.ac, data/00Index, data/Machines.rda, data/Oxboys.rda, man/Machines.Rd, man/Oxboys.Rd, man/corrmatrix-class.Rd, man/pdCompSymm-class.Rd, man/pdDiag-class.Rd, man/pdIdent-class.Rd, man/pdLogChol-class.Rd, man/pdMat-class.Rd, man/pdfactor-class.Rd, man/pdmatrix-class.Rd, src/Makevars, src/Makevars.in, src/lapack.h, src/pdMat.c, tests/pdCompSymm.R, tests/pdDiag.R, tests/pdIdent.R:
    New file.

    * DESCRIPTION, INDEX, R/AllClass.R, R/AllGeneric.R, R/pdMat.R, R/zzz.R, aclocal.m4, cleanup, configure.ac, data/00Index, data/Machines.rda, data/Oxboys.rda, man/Machines.Rd, man/Oxboys.Rd, man/corrmatrix-class.Rd, man/pdCompSymm-class.Rd, man/pdDiag-class.Rd, man/pdIdent-class.Rd, man/pdLogChol-class.Rd, man/pdMat-class.Rd, man/pdfactor-class.Rd, man/pdmatrix-class.Rd, src/Makevars, src/Makevars.in, src/lapack.h, src/pdMat.c, tests/pdCompSymm.R, tests/pdDiag.R, tests/pdIdent.R:
    Sources for R libraries