2026-11-19 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * R/simmer: sim() for merMod batches its linear algebra into one matrix
        product and builds each random-effects array once, instead of three
        matvecs and a split() per draw and an arr[i,,] <- that copied the whole
        array on every draw -- quadratic in n.sims, and the heavy garbage that
        made the display() example report over 5s CPU in R CMD check.  The
        draws stay bit-identical: rgamma() is rejection-based and consumes a
        variable number of uniforms, so they are still drawn one at a time and
        the caller's random stream is untouched.
      * R/extractDIC: dic.merMod evaluates the GLMM deviance over the whole
        (n obs x n.sims) matrix of draws instead of apply()ing family$aic()
        one draw at a time, reproducing family$aic() exactly including the
        binomial weights-into-trials folding.
      * NAMESPACE: import dbinom and dpois from stats.

2026-11-16 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * R/binnedplot, man/binnedplot: the plus and minus 2 SE band is the
        model-based 2*sqrt(p*(1-p)/n) of Gelman and Hill (2007, section 5.6) by
        default; se = "empirical" keeps the old band, which collapsed in
        low-probability bins because it used the sd of the residuals in the bin.
        The help example moves to the response scale (issue #53).
      * R/bayesglm: the Warning argument is read at last.  It was accepted,
        passed down and documented as controlling the non-convergence and
        separation messages, and was ignored.
      * R/bayesglm: the printed unnormalized log posterior passed the prior mean
        to dt()'s non-centrality argument, ignored the prior scale and used
        invlogit() whatever the link, and its log likelihood ignored the binomial
        totals.  It now uses the scaled-t prior and the family's link, and matches
        logLik().
      * R/bayesglm, R/model.matrixBayes: predict(type = "terms", newdata = ) no
        longer fails with "invalid 'labels'".  model.matrixBayes() copied
        attr(object, "intercept") onto its terms and a fitted model has none, so
        it copied NULL and the design matrix came back without an intercept.
      * R/bayesglm: a rank-deficient fit with a prior too flat to carry weight no
        longer fails with "object of type 'closure' is not subsettable".  The prior
        is doing nothing in that case, so the fit is handed to glm.fit(), which
        returns NA for the aliased coefficient and numbers for the rest.
      * R/bayesglm: a start of the wrong length reports the length it wanted
        instead of "the condition has length > 1".
      * R/bayesglm: with scaled = TRUE the prior scales, the augmentation means
        and the dispersion update are computed with the prior weights, so a weight
        of 2 now means the same as a duplicated row.
      * R/bayespolr: start is read as the documented c(coefficients, zeta) instead
        of in the optimiser's log-gap parameterisation, and a zeta that does not
        increase is rejected.
      * R/bayespolr: maxit bounds the optim() run as well as the start-value fit,
        and the cloglog start values come from a cloglog fit rather than a probit
        one.
      * R/sim: sim() dispatches on plm and coxph objects.  sim.plm() no longer
        multiplied the already-scaled vcov(<plm>) by sigma^2, and sim(lm) keeps
        aliased coefficients as 0 the way sim(glm) does.
      * R/balance: the raw sd-ratio column is computed rather than left NA, a
        matrix without column names no longer errors, and ... reaches the plot
        when plot = TRUE.
      * R/display: the unreachable .Call("mer_sigma") branch is gone, and an
        intercept-only lmer keeps its "(Intercept)" coefficient name.
      * R/rescale: a constant is centred instead of returning NaN.
      * R/dwish: coerces W, not S.
      * R/go, man/go.Rd: document the .R extension the function appends.
      * R/coefplot: any CI draws coefficients plus and minus CI sds, not just
        plus and minus 1 sd; the help page says so.
      * R/discrete.histogram: handles a single distinct value, a factor with an
        unused level, and ylab = passed through the dots.
      * R/triangleplot: the legend digit search no longer stops with "Too many
        digits!" on close-together correlations.
      * R/traceplot: a bugs trace plot labels its vertical axis "Kept draws", the
        index the axis actually carries, rather than "iteration".
      * man/multicomp.plot.Rd: documents abbreviate(label, minlength = 2) rather
        than "the first 2 characters".
      * tests/: 23 new regression-test files, one per finding above; each fails on
        the previous master with its own message.
      * R/bayesglm, R/model.matrixBayes: a formula whose term is a transformation
        of a variable -- poly(psi, ...), log(x) -- no longer fails with "object
        'psi' not found" when the fit also has weights.  model.matrixBayes() used
        to subset the model frame it was handed, which drops the frame's terms
        attribute and the columns the terms refer to (a frame holds the evaluated
        poly() column, not psi).  Found by the reverse-dependency check of 1.15-4:
        qgcomp's bayes path failed on it.
      * R/model.matrixBayes, R/bayesglm: model.matrixBayes(formula, data) still
        returns no intercept column -- a formula carries no intercept attribute to
        copy -- because FindIt prepends its own intercept and broke when that
        changed.  predict(type = "terms", newdata = ) is fixed in its caller
        instead, which now builds the contributions from the fit's terms and model
        frame.
      * tests/: two more regression-test files for those, and a test for the
        shape model.matrixBayes(formula, data) returns.
      * DESCRIPTION: the license is GPL (>= 2), as in every published release.
        The 1.15-4 tree had moved it to GPL (>= 3) on 2026-11-07 (entry below);
        that line is withdrawn.  GPL-3 is one-way incompatible with anything that
        still has to work with arm on GPL-2 terms, and 12 of the 44 reverse
        dependencies declare GPL-2 with no later-version option: BayesFactor,
        blmeco, brms, carcass, factorEx, ltmle, medflex, mlVAR, PSPI, SBMTrees,
        SEIRfansy, SynDI.  A relicensing is not part of a bug-fix release.

2026-11-15 Gaurav Sood     (correctness audit; index in issue #54)
      * R/bayespolr: scaled = TRUE now rescales the prior as documented
        (prior.scale/range(x) for a two-valued predictor, prior.scale/(2*sd(x))
        otherwise).  The block was guarded by xint > 1, where xint is the column
        of the intercept, so it never ran (pull request #29).
      * R/bayespolr: deviance (and so logLik and AIC) is that of the data.  It was
        twice the optimised objective, which also holds the log prior and the
        prior pseudo-counts.  Fitted probabilities now include the offset, as in
        MASS::polr (pull request #30).
      * R/bayesglm: for a gaussian response the x-scaling of the prior is kept on
        every iteration.  It was undone after the first, so estimates depended on
        the units of x (pull request #46).
      * R/bayesglm: AIC and logLik for a binomial response with more than one
        trial pass the trials vector to family$aic, as glm.fit does (pull request #40).
      * R/bayesglm: the design matrix is built from the model frame, so subset,
        na.action and drop.unused.levels are honoured (pull request #41).
      * R/bayesglm: the stored terms are built with keep.order, so predict(se.fit =
        TRUE) and type = "terms" pair columns with the right coefficients (pull
        request #44).
      * R/bayesglm: the dispersion update for families other than binomial and
        poisson uses only rows with positive weight; with scaled = FALSE a zero
        weight equals dropping the row (pull request #42).
      * R/bayesglm: an offset combined with a vector prior no longer errors in the
        null-deviance refit (pull request #43).
      * R/bayesglm: control = list(maxit = ) is honoured when maxit is not given
        (pull request #45).
      * R/extractDIC, R/display: DIC for merMod fits is the conditional DIC computed
        from sim() draws, pD = mean deviance - deviance at the mean draw.  The old
        pD was 0 for ML lmer fits and negative for glmer fits.  extractDIC() gains
        n.sims and returns pD as an attribute; display() uses a fixed internal seed
        and restores the caller's (pull request #52).
      * R/display: for glmer fits without a scale parameter, no Residual sd row
        (pull request #51); for svyglm, n is the number of observations, not the
        design degrees of freedom (pull request #50).
      * R/balance: the matched weights are applied to treated units as well as
        controls, which matters for IPTW, ATE and ATC (pull request #35).
        plot.balance() subsets longcovnames with which.covs (pull request #36);
        print(combined = TRUE) reads diff.means.matched (pull request #37).
      * R/matching: replace = FALSE stops when there are fewer controls than
        treated units, instead of returning NA matches (pull request #38).
      * R/standardize: the data and variables are found in the environment where
        the model was fitted, not in arm's namespace or the global environment
        (pull request #39).
      * R/standardize: stops with an explanation when rescaling makes a transformed
        input undefined (e.g. log() of a centred variable), instead of silently
        dropping rows; standardize() no longer errors when nothing is rescaled
        (pull request #55).
      * R/sim: n.sims = 1 for glm no longer returns zeros; an intercept-only polr
        no longer puts a cutpoint into the coef slot (pull request #48).
      * R/sigma.hat: sigma.hat() for a glm returns sqrt(dispersion) for every
        family; it returned NULL for all but gaussian (pull request #47).
      * R/se.coef: se.coef() and se.ranef() handle a list-valued postVar, as for
        two random-effect terms on one grouping factor (pull request #49).
      * R/coefplot: the axis range covers user-supplied confidence bounds, which
        also no longer require sds (pull request #31); the bugs method honours
        pch.pts (pull request #32).
      * R/binnedplot: binned.resids() returns the bin breakpoints in xbreaks
        (pull request #33).
      * man: coefplot, binnedplot, matching, contrasts.bayes, mcsamp and bayesglm
        corrected where they disagreed with the code (pull request #34).
      * DESCRIPTION: URL lists the GitHub repository (pull request #28).
      * tests: a regression test for each code change above (all except #28 and
        #34), each checked to fail on the code it guards.

2026-11-07 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.15-4
      * R/rescale: "binary.inputs = -0.5,0.5" now maps the two observed values onto -0.5
        and 0.5 whatever the variable is coded as.  It previously subtracted 0.5 from x,
        which is only correct for a variable already coded 0/1, so a numeric binary
        variable coded 1/2 gave 0.5 and 1.5.  That is the same bug as issue #19, which
        had been fixed for factors only.
      * R/rescale: an unrecognised "binary.inputs" is now an error.  For a binary
        variable a value outside the four documented options fell through every
        branch and returned NULL invisibly, and for a continuous variable it was
        silently ignored.
      * man/rescale: state the binary.inputs options precisely, including that "0/1" and
        "-0.5,0.5" are invariant to how the variable is coded.
      * tests: first regression suite, run by R CMD check.  Plain R scripts on purpose,
        so there is no new dependency.  tests/simmer.R pins the sim() covariance fix
        (issue #26), tests/load-first.R the .onAttach guard (issue #25),
        tests/standardize.R the formula-in-an-object fix (issue #24) and tests/rescale.R
        the binary handling above.  Each was checked to FAIL against the code it guards,
        not merely to pass against the fix.
      * README.md: rewritten.  It was a title and a single line; it now describes
        what the package is for, how to install it, the parts most people want
        (bayesglm, display, standardize/rescale, sim, the plots, matching) with
        their output, two example figures drawn from the shipped lalonde data,
        and a grouped function index.

2026-11-07 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: license changed from GPL (>= 2) to GPL (>= 3), i.e. GPL-3.0-or-later.
      * R/simmer: fix sim() for merMod, whose random-effect draws did not match the model's
        covariance: the fill-reducing permutation of the sparse Cholesky factor was never
        undone, and Lambda was transposed (issue #26; fix contributed by David Skalinder
        in pull request #27).
      * R/simmer: undo that permutation with a sparse permutation matrix instead of by
        base-R subassignment, the latter being several hundred times slower on large
        models (follow-up to pull request #27).
      * R/load.first: .onAttach no longer derives the library location from system.file(),
        which searches .libPaths() rather than the library the package was loaded from;
        it now uses the libname R passes to the hook, and no longer aborts when the
        package description cannot be found (issue #25).
      * NAMESPACE: import sparseMatrix from Matrix.

2026-04-15 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.15.2
      * R/standardize: fix a bug when formula is inputed as an object.
      * man/standardize: and an example to reflect the aformentioned change.

2026-03-15 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.15.2
      * R/rescale: fix a bug previous coded in rescale that breaks standardize()

2026-03-15 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.15.1
      * R/sim: add hidden function for sim.plm and sim.coxph
      * R/standardize: rewrite the function
      * R/rescale: rewrite rescale function to correclty handly binary varaible.

2022-8-25 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.13.1
      * NAMESPACE: import solve from Matrix
      * R/simmer: comment out solveFun(), and use solve from Matrix direclty

2021-10-15 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.12-2
      * man/lalonde.Rd: fixed the link to lalonde data (from http to https)
     
2021-10-08 Marius Barth   <marius.barth@uni-koeln.de>
      * DESCRIPTION: (Version, Date): removed the Hmisc package from Imports field
      * NAMESPACE: Do not import wtd.var from Hmisc, anymore (this is to increase crossplatform compatibility)
      * R/balance.R: Replace call to Hmisc::wtd.var() with call to stats::cov.wt()


2020-7-27 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.11-2
      * NAMESPACE: import weighted.mean from stats and wtd.var from Hmisc
      * R/balance: new balance, print.balance, plot.balance function
      * R/matching: new matching function
      * man/balance: new description to new functions

2020-4-27 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.11-0
      * NAMESPACE: import setClass from methods  (BUGS reported by Henrik)

2018-4-12 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.10-1
      * R/bayesglm: fix a bug where scale=TRUE the prior.scale miscount the nvars.
      * R/sim.glm: improve the speed.
      * man/standardized: fix a typo in the example

2016-11-24 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: (Version, Date): 1.9-3
      * DESCRIPTION: new description, and change http to https in URL

2016-9-4 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: add BugReports and change LICENSE to GPL >=3, fix Vincent's name (spelling error, sorry Vincent!)
=======
2016-9-2 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
      * DESCRIPTION: add BugReports and change LICENSE to GPL >=3
      * Doc fix Vincent's name

2016-8-21 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.9-1
   * NAMESPACE: import show from methods


2015-7-7 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.8-6
   * NAMESPACE: import more from base packages (new R rules)

2015-5-3 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.8-5
   * R/bayesglm: fix a missing line in the re-factorization of bayesglm.fit, when scaled=TRUE, and a column of X takes on more than two values, than prior.scale = prior.scale /(sd(x)*2).

2015-4-7 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.8-4
   * R/sigma.hat: sigma.hat changed to S3 function.
   * R/fitted: changed to S3 function
   * R/coef: coef, fixef, ranef, changed to S3 function


2015-3-31 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.8-03
   * R/bayesglm: revert back to a more straightforward coding, easier for debugging.
   * R/bayespolr: check n.iter to maxit, and pass it through control.


2014-8-27 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.7-07
   * R/readColumns: add read.columns()
   * R/sim: fix a bug in the name calling of beta.hat
   * man/readColumns: add description for read.columns()
   * NAMESPACE: export read.columns
   * DESCRIPTION: remove foreign and R2WinBUGS from suggests



2014-8-1 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.7-05 (thanks to Dr. Martyn Plummer's contribution)
   * R/bayesglm: fix several bugs in bayesglm
   * R/display: change the display for bayesglm to fit the changes stated above.
   * man/bayesglm:
        1. change M2 and M7 example codes to make them equivalent to M1 and M3.
        2. change the description for prior.scale for the gaussian family.
   * DESCRIPTION: remove foreign and R2WinBUGS from suggests

2014-4-27 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.7-03
   * R/bayesglm: revert back to the use of lm.fit to lm.wfit; put a stop when dispersion goes Inf

2014-4-24 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.7-02
   * R/simmer.R: simmer attaches names to fixed effects
   * R/fitted: fitted uses correct observational weights for glmms
   * R/bayesglm: use change the use of lm.fit to lm.wfit.

2013-11-25 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-10
   * make changes that fit to oldrelease R.
   * R/se.ranef, se.coef: change postVar to condVar

2013-9-23 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-09
   * start supporting new lme4

2013-9-23 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-09
   * start supporting new lme4

2013-8-22 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-07
   * revert back to 1.6-07, stop supporting the new lme4.
   * R/bayesglm: fix various bugs

2013-8-21 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-08
   * clean up DESCRIPTION and NAMESPACE

2013-7-12 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-07
   * made various change to adjust the new lme4
   * currently, sim.mer is not working, waiting for revision.


2013-5-9 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-06
   * NAMESPACE: export new method fitted() for sim.mer
   * R/fitted: add fitted() for sim.mer object
   * man/sim: add description for fitted() for sim.mer object
   * R/bayesglm fix a bug in bayesglm() in "subset"

2013-3-8 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-05
   * NAMESPACE: export logit()


2013-2-27 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-04
    * R/coefplot: fix a bug when the formula does not have an intercept
   * man/bayesglm: fix a coding error for weights in the bayesglm.fit


2013-2-20 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-03, add import from survey package
   * R/AllClass: set old class svyglm.
   * R/display: add svyglm method
   * R/coefplot: fix a bug when the formula does not have an intercept
   * man/display: add svyglm method

2013-1-5 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.6-01
   * R/bayesglm: fix an issue in updating start in the loop

2012-10-13 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * R/coefplot: fix an issue in coefplot.  No longer reset par when exit.
   * R/bayesglm: fix an issue in dev and family$state$valideta, family$state$mu

2012-10-03 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.5-08
   * R/bayesglm: fix various bugs in bayesglm

2012-09-26 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.5-07
   * R/balance: handle the situation when the formula in pscore.fit is not directly express.

2012-09-20 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.5-06
   * R/bayesglm: stop using .Fortran() here.

2012-06-6 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.5-05
   * man/bayesglm: add predictLM
   * R/bayesglm: add predict.bayesglm, predictLM to fit with model.matrixBayes

2012-04-27 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.5-04
   * man/bayesglm: rewrite the description for the option scaled.

2012-03-3 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.5-03
   * DESCRIPTION: add foreign as the required package
   * R/simmer: new sim() for mer class

2012-01-19 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.5-01
   * R/mcsamp: add mcsamp() back, though it is not working.
   * R/AllGeneric: set coef, print, as generic to pass the check

2011-11-21 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-14
   * R/.onAttach: fix the NOTE issue

2011-06-19 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-13
   * R/coefplot: fix margin control.

2011-06-11 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-12
   * R/bayespolr: add min.prior.scale
   * R/binnedplot: add nclass > 1 check
   * man/bayespolr: add min.prior.scale


2011-05-25 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-11
   * R/bayesglm: fix a bug  when there are some observation-weights that are zero.

2011-05-9 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-10
   * R/display: fix a bug in display.lm that fails to print out se.


2011-05-8 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-09
   * R/binnedplot: fix a bug of no sd when binnedplot only get one point.

2011-04-24 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-08
   * display: now return objects after displaying the fitted model.
   * sigma.hat: fix a bug in sigma.hat for mer


2011-03-1 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-07
   * AllClass:  add sim.polr class
   * coef: add coef() for sim.polr
   * sim: add sim() for polr

2011-03-1 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-06
   * NAMESPACE: export distcrete.histogram

2011-02-23 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-04
   * R/binnedplot: pass addition graphical parameters to the function.

2011-02-15 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-03
   * R/sim: fix a bug in sim.glm() when there is only an intercept
        as a predictor. (discovered by Barnes Benjamin)


2011-02-14 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-02
   * R/bayesglm: fix some dimension issues when NCOL(x.nobs)==1

2011-02-05 Yu-Sung Su     <suyusung@tsinghua.edu.cn>
   * DESCRIPTION (Version, Date): 1.4-01
   * R/load.first: lib --> lib.loc in packageDescription

2010-11-20 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-08
    * R/simmer: samples directly from the posterior
        of the fixed and random effects, given sigma and Sigma

2010-10-24 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-07
    add new methods for sim object, coef, fixef, ranef and sigm.hat

2010-9-24 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-06
    * R/bayesglm: fix a bug when a model of one predictor with no intercept is fitted
    * man/several: CRAN no longer alows genericFunction docType

2010-6-28 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-05
    * R/extractDIC: add s3 methods for extractDIC and extractAIC for the mer class
    * Rd/extractDIC

2010-1-21 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-04
    * R/standardize: add polr method
    * Rd/standardize: change the example code to make M1 and M2 equivalent.

2010-1-15 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-03
    * R/balanceplot: fix a bug in balance(), take out the intercept

2010-1-11 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-02
    * R/bayesglm: new bayesglm.fit (written by Daniel Lee)


2010-1-8 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.3-01.
    * R/bayesglm: a bug in x.matrix augmentation


2009-12-30 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-13.
    * R/bayesglm: smarter use of x matrix to save memory usage


2009-12-12 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-12.
    * R/simmer: use of sparse matrix in sim.mer

2009-12-08 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-11; add abind pacakge dependencyR
    * R/simmer: new sim functions for "mer" class

2009-11-22 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-10.
    * NAMESPACE: export logit

2009-4-29 Yu-Sung Su     <ys463@columbia.edu>
    * R/display: fix format inconsistency in sprintf (fround)

2009-4-13 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-9.
    * R/macthing: fix a bug in matching replace=TRUE


2009-3-31 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-8.
    * R/display: add option detail


2009-3-1 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-7.
    * R/display: fix a bug in display.mer


2009-2-26 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-6.
    * R/display: print out t-value, z-value, p-value

2009-2-20 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-5.
    * R/load.first: no longer set default digit=2

2009-2-17 Yu-Sung Su     <ys463@columbia.edu>
    * NAMESPACE: export binned.resids


2009-2-4 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-4, no longer need car
    * R/load.first: car is no longer required

2009-2-1 Yu-Sung Su     <ys463@columbia.edu>
    * man/coefplot: fixed doc error
    * man/sim: fixed doc error


2009-1-30 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-3.
    * R/coefplot: fixed margin bugs in coefplot.default

2009-1-29 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-2.
    * R/sim: for mer method, add option ranef. Users choose to return
      sim.ranef or not.

2009-1-28 Yu-Sung Su     <ys463@columbia.edu>
    * man: first attempt to clean up help files to comply the new rule.
            in particular, use \dQuote and \sQuote for "" and ''.
    * man/bayesglm: update reference
    * man/bayespolr: update reference

2009-1-22 Yu-Sung Su     <ys463@columbia.edu>
    * man/rescale: update reference.
    * man/standardize: update reference.

2009-1-16 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-1.
    * NAMESPACE: export S4 method for standardize
    * man/standardize: add description for S4 methods.
    * R/AllGeneric: add a generic function for standardize
    * R/standardize: add standardize.default, S4 methods for lm, glm and mer.

2009-1-03 Yu-Sung Su     <ys463@columbia.edu>
    * DESCRIPTION (Version, Date): 1.2-0.
