Release 0.95-8 with bug fix for sparse symmetric solve

More logical sparse matrix methods

Getting ready for a release

Added colSums and friends for dgeMatrix class

Updates related to Laplace approximation

changes related to Laplace approximation, including new function to compute parts of the deviance in C code, which doesn't currently work

additions related to the Laplace approximation in glmm lmer models

-Wall

Give a warning, not an error, for detection of exact singularity in the LU decomposition

Release 0.95-6

no need for \ in %*% in error message

Create "fallback" methods for any kind of numeric, dense matrix

Moving methods for operations with numeric to the Matrix class and with integer to the dMatrix class

Repaired a thinko in the lmer_fitted function

Rename expand_column_pointers to expand_cmpPt (expand compressed pointers)

Added dgRMatrix methods and documentation

export csc_tcrossprod - and add examples & test for tcrossprod()

eliminate extraneous declaration for "-Wall"

Generalize matrix multiplication and solve methods

include permutation in expand method for LU

Updates to multiplication and other methods

Validate function stubs for pBunchKaufman and pCholesky classes

Zero the unused triangle in the Cholesky decomposition

Adding many methods for the dsyMatrix class

Coercions and methods for dppMatrix and dspMatrix classes

Coercions and methods for dppMatrix and dspMatrix classes

Coercions and methods for dppMatrix and dspMatrix classes

comments

Add BunchKaufman factorization

Fix dgeMatrix_dtpMatrix_mm and add test

more  %*% methods; some dimnames in C (untested); checks extended

Wrong number of args in lmer_update_mm

L <-> U for transpose

+ Matrix_DimnamesSym

do not Calloc()/Free() in "good" case of check_scalar_string()

check for integer 'x'

Define %*% method for dtpMatrix and matrix properly

Added C function lmer_fitted

Methods and coercions for the dtpMatrix class

Clean up *_validate functions

C functions for dtpMatrix class

more coercions

shorten csc_components_transpose to csc_compTr

dsTMatrix_as_dsCMatrix

more "-Wall" dummy init.s

typo in SelectQueue[o]neWay; comment

Internationalization and localization

prevent segfault {but not "nicely"}

use "(.)" against "-Wall" warning

Inlined function declarations moved within the conditional expression

Clean up documentation errors exposed by doxygen

Fix compiler warnings

Clean up warnings from compiler

Argument list for cscb_trsm

Changes for non-trivial cscb-trsm

Coercions and methods for the dsTMatrix class

Non-trivial cscb_ldl case

cosmetic changes - reformating

clean up logic

clean up long lines

triangular, triplet sparse matrices

Methods and coercions for the dtTMatrix class

Changes for non-nested vector-valued random effects

rearranging utility functions in the C code

use mkString(); "uplo" and "diag" must be uppercase

Include the $(SUBSTAMP) target

Include a dependence of the LIB target on the OBJS

Force a call to Make lib in the subdirectories whenever the SHLIB
target is built.  Right now this isn't working properly.

- get rid of two extraneous declarations
- use the SWAP(.) macro {as in other places}

Initial implementation of the Schur generic

Move the make_named function to Mutils as Matrix_make_named

Getting package to pass CMD check

renaming and rearranging classes

renaming and rearranging classes

renaming and rearranging classes

renaming and rearranging classes

renaming and rearranging classes

Renaming and reorganizing classes

Renaming and reorganizing classes

Change name of slot from factorization to factors

Add an externally accessible entry point to lmer_Crosstab

Add an externally accessible entry point to lmer_Crosstab

Clean up Lind calculation

Make make clean cleaner

Add expm function

New coercion method for cscBlocked objects

Release 0.8-23

Change char flags to cblas enum types

Off-diagonals from L, not ZZpO

Off-diagonals from L, not ZZpO

Remove references to the Linv slot.  Change to the extended parent
pairs.  Extensive changes in lmer_invert.  These still probably have
errors but this check in is to provide a checkpoint.

Remove any references to Linv.  Rewrite routines to use a solution
from L rather than multiplication by the inverse.  Change calling
sequences to use the Parent array.  Change the parent arrays to
extended parent pairs.

See if use of cblas would be a good idea

See if use of cblas would be a good idea

Remove mistaken transpose

Remove Linv references

Added and modified many functions to do solutions rather than inversions.

Define the AZERO macro

Use the AZERO macro

Add definitions of cscb_trsm and cscb_trcbsm

Many changes with respect to the representation and calculation steps.

decrement when counting down

Minor typos

Check in prior to leaving for Zuerich

Change to the lmer representation

Change to the lmer representation

Working on the lmer classes

Add arguments to ssclme_to_lme

Add a coercion for tripletMatrix to matrix

Keep -Wall happy

Fix bug in multivariate/multiple groups case

make -Wall happy

make -Wall happy

make -Wall happy

declaration of R_alloc

Fixed nested scalar r.e., multivariate bug still outstanding

R-specific versions of ldl

R-specific versions of ldl

R-specific versions of ldl

R-specific versions of ldl

R-specific versions of ldl

R-specific versions of ldl

R-specific versions of ldl

Add a call to ldl_numeric to obtain row indices of diagonal blocks of L

separate suboperations from Parent_inverse

silly typo

Add block updates in bCrosstab_convert

Make Parent_inverse visible in include file

Extract dependencies to an include file

Move include file to C sources so the .h file can be included without pulling in a bunch of other stuff

add blocked crosstabulations

keep valgrind happy

Fix on uninitialized value

Updates for factorization name and default value

Use Matrix_rcondSym

Add syrk code

Add syrk code

Initialize all slots

initialize all slots in a NEW_OBJECT

Keep symbols and filenames consistent

Keep Makefile.win in sync with Makefile

Keep -Wall happy

add a blas-like routine

add a blas-like routine

cosmetic changes

Changes for check on Debian dgesdd

Changes for check on Debian dgesdd

updates and downdates

lmeRep_invert now applies to multiple, nested grouping factors

transfer to another computer

Adding lmeRep_L_sm and support routines

Only positions, not counts, from crossproduct

added validate functions

Add registration of .Call functions

Add registration of .Call functions

Release 0.8-15

Bug fixes in 0.8-14

minor format change

New storage format - currently works for single grouping factor

Interim version of lmeRep - not yet complete

Add cscBlocked, tweak lmeRep

New version

Initial lmeRep_update_mm modifications for multiple grouping factors

extending lmeRep to multiple nested grouping factors

Remove taucs, add lmeRep, new release

Add the lmeRep class

cleanup in EMsteps_verbose_print

Combine ssclme_coef and ssclme_coefUnc, ssclme_coefGets and ssclme_coefGetsUnc.  Reorganize gradient code in preparation for Hessian calculations.

Remove warnings, new release, change gradient calculation method

Remove pdMat classes

Updated Hessian method to symmetrize the result

Added stub of a Hessian generic and methods.

Combine the two ldl_symbolic routines

Experiments with projecting

Adding documentation - cleaning up code

remove unused variable

Multiple factors with some ni > 1

Accurate EM iteration count

Moving utilities to Matrix package

Remove storage of L-inverse

maxLI=0 option works

More attempts at the bVar calculation

Add routines to project and ldl_symbolic the ctab

Confused conflicts in checkins

Previous to changing permutation method

Rearranging code to isolate Metis includes

Yet another permutation calculation method

Permutations and update_mm

Remove useLast argument

gradients missing when deviance missing

Change ssclme_variances call

updated ssclme_collapse

Added ssclme_collapse

Correction for gradient and EMsteps

Release 0.8-5

Logic error in ssclme_update_mm

More work on gradient (still not correct)

Added gradient, consistent order of parameters

analytic gradient except for unconstrained, multidimensional case

Add gradient calculation

First cut at gradient

Cleanup of header file

New method of choosing between ties

Allow useLast option in sscCrosstab_groupedPerm

Working version of sscCrosstab_groupedPerm but about to do surgery on it

Working version of sscCrosstab_groupedPerm - still needs work

Updates to sscCrosstab_groupedPerm

New (experimental) groupedPerm

ssclme_EMsteps miscalculation

Fix Makefile.win

Original version of sscCrosstab_groupedPerm is about to be dismembered

R version of fill-reducing permutation

Release 0.8-3 to fix Makefile.win

Adding coercions - fixing bugs

Minor change for VarCorr

Lack of inclusion picked up by -Wall

Removing explicit umfpack code

Move triplet_to_col from umfpack

Added ssclme_variances

remove unnecessarily complicated C code for coercion (better done in R code)

Correct column of b[[i]] in _fitted

Added ssclme_fitted function

Dimensions of b-related values, dimnames

Changing dimensions in bVar and returned value from _ranef

Allow unconstrained coefficients

Change name of C function ssclme_EMsteps

Initial import