0.4.1 fixed documentation for metrop
0.5 changed licence to X11 (was AFL, which apparently sucks)
changed to NAMESPACE
cleaned up latex in inst/doc, added Makefile
removed src/Makevars
had to get rid of R_IsNaNorNA in src/metrop.c
generally, lots of changes to get ready for CRAN
all of the tests/*.Rout.save redone because the digits
in numerical printing seem to have changed -- cannot get
bit for bit identical with earlier R (even before code changes).
changed first arg of metrop from "o" to "obj"
0.5-1 changed package vignette to fix bogus MCSE calculation
(bogosity still admitted and explained in appendix).
0.6 added initseq function
0.7 added temper function -- not tested yet
0.7-1 fix bug && to ||
much confusion about whether my_i and my_j were 0-origin or 1-origin
forgot to allocate debug_acceptd
confusion about dimension of state (nx + 1 vector if serial,
ncomp x nx matrix if parallel) and <= nx versus < nx in loop bounds
bug where my_swapped_proposal_log_dens was used twice when one should
have been my_swapped_coproposal_log_dens
bug in dimension of debug_state when parallel (last dim nx not nx + 1)
accepted swap didn't actually swap, fixed
all the accept decisions when log_hastings_ratio < 0 were backwards
batch means accumulation has = instead of +=
batch means divided by nbatch rather than blen
accepti entries stored in wrong places in output structure
ripped out old system for caching lud values, put in new that caches all
(more efficient, but real reason was old was broken somehow)
of course, its first implementation had 0-origin vs 1-origin confuse
yet another instance 0-origin vs 1-origin confusion (my_i should be
my_i - 1
length of output structure (len_result) determined incorrectly, fixed
caching of current lud value broken in serial case (stored in i not j)
after all that, everything now checked even with --use-valgrind and
--use-gct
0.7-3 add bugs section to initseq.Rd
fix example in temper.Rd to actually be for temper rather than metrop
fix comment in temper.c
add inst/doc/debug.Rnw a vignette about debugging and debug output
0.7-4 some changes to /inst/doc/demo.Rnw
0.7-5 add inst/doc/bfst.Rnw a vignette about Bayes factors and serial tempering
0.8 earlier versions calculated the Hastings ratio wrong in doing serial
tempering, ignoring the number of neighbors each component had, hence
computing wrong answers when components differed in number of neighbors
0.9 morph, morph.metrop functions and morph.pdf vignette
bumped required R version to 2.10.0 (required by xtable)
0.9-1 added PACKAGE = "mcmc" where missing to .C and .Call
0.9-3 move vignettes to vignettes directory
changed LICENCE file yet again as per CRAN requirements
0.9-4 cleaned up some tests
import from stats, as required by R-3.3.0
0.9-5 use registration of native routines described in Sections 5.4 and
6.15 of Writing R Extensions
add calls to cmpfun inside metrop and temper
0.9-6 take out calls to cmpfun and requirement for R package compiler
(R now automagically does jit)
Fix problems found in CRAN static analysis
0.9-7 add calls to R_CheckUserInterrupt in C functions mcmc and temper
Generated by dwww version 1.15 on Wed Jun 26 01:25:30 CEST 2024.