Movatterモバイル変換

NEWS	R Documentation

lme4 News

CHANGES IN VERSION 1.1-38

BUG FIXES

$se.fit component of predictions from GLMMs whentype = "response" is now transformed from the link scale tothe response scale (GH #847, @leorjorge)
predictions withse.fit=TRUE/allow.new.levels=TRUE work for a wider range of cases (GH #691, GH #753)

NEW FEATURES

thestart argument to[g]lmer now allows'beta' as a synonym for 'fixef'
confint.merMod now allowssignames as anargument instead ofoldNames, for consistency withprofile.merMod;oldNames is deprecated
added six publically available datasets:
- cbpp2 (contagious bovine pleuropneumonia (extended version ofcbpp)),
- culcitalogreg (coral-eating seastarCulcita novaeguineae). Also added the alternative version,culcitavolume,
- gopherdat2 (gopher tortoises shell remains),
- salamander (mountain dusky salamander mating),
- schizophrenia (National Institute of Mental Health schizophrenia collaborative study),
- toenail (toenail onychomycosis data from dermatophyte infections).
allFit(..., start_from_mle = TRUE) now specifies that models should be refittedusing starting parameter values from the original model fit,which should increase efficiency
Predictor variables can be now automatically scaled and centered via(g)lmerControl(..., autoscale = TRUE), with resultsreported on the original data scale. Seevignette("autoscale", package = "lme4")for details (wish of GH #227, although at present only predictorvariables can be scaled)

USER-VISIBLE CHANGES

lme4 now depends on thereformulaspackage for formula processing etc.; the corresponding functions(expandDoubleVerts,subbars,findbars,nobars, etc.) are still (re-)exported from the package,but throw a warning if called directly(use e.g.reformulas::findbars(...) to avoid the warning)
derivative-based convergence checks are now skipped formodels with large data (number of observations greater thanthecheck.conv.nobsmax component of[g]lmerControl()); default threshold is 1e4 observations.If desired, derivatives can be calculated (but not checked)by explicitly settingcalc.derivs to TRUE.
Derivative calculations are now skipped for models withsingular random effects components (unlesscalc.derivs isexplicitly set to TRUE in[g]lmerControl).

CHANGES IN VERSION 1.1-37 (2025-03-26)

NEW FEATURES

simulate.merMod and.simulateFun gain thecluster.rand argument to allow specifying non-Normalcluster random effects in simulations.
simulation.Rmd, and derived .md and .html files inmisc/notes/, trace the flow of the simulation just beforethis change. Likely of interest only to developers.

USER-VISIBLE CHANGES

headline ofprint andsummary output forglmer fits now labels the minimum of the objective function(correctly) as "-2*log(L)" rather than "deviance" (inspired byglmmTMB GH #1156)

CHANGES IN VERSION 1.1-36 (2025-1-11)

NEW FEATURES

The full (joint) conditional covariance matrix of fixedeffects and conditional modes is now available viavcov(fitted_model, full = TRUE)
Random effects formulas are now processed by code fromthereformulas package, meaning that someextended random effects formulas (such as(1|f*g) forcrossed random effects) will now work

USER-VISIBLE CHANGES

simulate withnewparams no longer printsa message if the individual parameter vectors are unnamed

BUG FIXES

floating-point errors leading to slightly negative deviancescould lead to NaN deviance residuals (Wolfgang Viechtbauer, GH #812)

CHANGES IN VERSION 1.1-35.5 (2024-07-03)

USER-VISIBLE CHANGES

inpredict, the synonymsReForm,REForm, andREform ofre.form are now hard-deprecated,giving an error (after 10 years of soft deprecation)

OTHER CHANGES

minor adjustments in test tolerances
correct Matrix dependency to >= 1.2-3

CHANGES IN VERSION 1.1-35.4 (2024-06-19)

BUG FIXES

predict(., re.form=...) works in a wider range ofcases (GH #691)
Gamma simulation now uses correct shape parameter (GH #782)
Avoid triggering RcppEigen UBSAN bug(?) in the case ofprofiling fixed effects in amerMod object witha single fixed effect parameter (GH #794: lots of help fromDirk Eddelbuettel and Mikael Jagan)
fix bug in plot methods (cbind'ing zero-length objects)

CHANGES IN VERSION 1.1-35.3 (2024-04-16)

BUG FIXES

bug-fix for ASAN/memory access problem in CholmodDecomposition(Mikael Jagan)

CHANGES IN VERSION 1.1-35.2 (2024-03-28)

This is primarily a 'bump' release to ensure that packagerepositories rebuild binaries with the latest version of theMatrix package.

BUG FIXES

simulate works (again) withre.form=NULL whenNA values are present in the data (GH #737, @frousseu)

USER-VISIBLE CHANGES

updated tests of upstreamMatrix version; shouldnow warn only on ABI incompatibility, not on package versionmismatch alone

CHANGES IN VERSION 1.1-35.1 (2023-11-05)

USER-VISIBLE CHANGES

lFormula andglFormula once againdoallow matrix-valued responses (for use in downstream packages likegalamm)

CHANGES IN VERSION 1.1-35 (2023-11-03)

NEW FEATURES

predict.merMod now has ase.fit method, whichcomputes the standard errors of the predictions, conditional on theestimatedtheta (variance-covariance) parameters

USER-VISIBLE CHANGES

usinglmer with a matrix-valued response now throwsa more informative error message, directing the user to?refit

CHANGES IN VERSION 1.1-34 (2023-07-04)

BUG FIXES

summary(<merMod>) now records ifcorrelation wasspecified explicitly and to what; and itsprint() method takes itinto account; notably summary(<merMod>, correlation=TRUE) will bydefault print the correlation (matrix of the fixed effects) fixing GH #725

NEW FEATURES

refit gains anewweights argument

CHANGES IN VERSION 1.1-33 (2023-04-25)

BUG FIXES

a boundary check could fail occasionally when large dataproduced an NA value in a computed gradient; now warns instead(GH #719, Mathias Ambuehl)
allFit now works better whenoptimx anddfoptim packages are not installed (GH #724)
refit reset internal degrees of freedom componentincorrectly for REML fits (resulted in incorrect reported REML criteria,but otherwise harmless: side effect of GH #678)

NEW FEATURES

dotplot andqqmath methods gain alevelargument to set the width of confidence intervals
dotplot method is now more flexible, using ".v"options (lty.v,col.line.v,lwd.v) to setappearance of vertical lines (Iago Giné Vázquez)
refit gains anewweights argument (GH #678)

CHANGES IN VERSION 1.1-32 (2023-03-14)

USER-VISIBLE CHANGES

formatVC() gets a newoptional argumentcorr indicating if correlations orcovariances should be used for vector random effects;this corresponds toprint(<merMod>, ranef.corr = ...) .By default, it is FALSE forcomp = "Variance", fixing (GH #707).
qqmath.merMod adds a (useless)data argumentfor S3 compatibility. Going forward, theid andidLabels arguments should always be specified by name.We have added code to try to detect/warn when this is not done.

BUG FIXES

nobars now retains the environment of its formulaargument (GH #713, Mikael Jagan)

CHANGES IN VERSION 1.1-31 (2022-11-01)

BUG FIXES

confint(fm, <single string>) now works (after years ofbeing broken) again.
simulating from binomial model with a factor response, whenthe simulated response contains only a single factor level, now works(Daniel Kennedy)

CHANGES IN VERSION 1.1-30 (2022-07-08)

USER-VISIBLE CHANGES

nl (term names) component added to output list ofmkReTrms (GH #679)
eliminate partial-matching ofeta (foretastart) (GH #686: not actually "user-visible" unlessgetOption("warnPartialMatchDollar") isTRUE)
summary method doesn't break for GLMMs other thanbinomial/Poisson whenmerDeriv'svcov.glmerMod method isattached (GH #688)

BUG FIXES

better handling ofsimulate(., re.form = NULL) whenmodel frame contains derived components (e.g.offset(),log(x))(https://github.com/florianhartig/DHARMa/issues/335)
bootMer works withglmmTMB again (broken in1.1-29)
maxfun argument toallFit controls maxfunction evaluations for every optimizer type (GH#685)

CHANGES IN VERSION 1.1-29 (2022-04-07)

USER-VISIBLE CHANGES

prediction with new levels (when not allowed) returns a moreinformative error message (displays a list of unobserved levels)

BUG FIXES

glmer.nb now works whenlme4 is not loaded (GH #658, @brgew)
tests for singularity (check.conv.singular) now runindependently of derivative computation (e.g., whencalc.derivs=FALSE) (GH #660, @palday)
influence.merMod now works when data were originallyspecified as a tibble
fixed bug incooks.distance method forinfluence.merMod (i.e., objects created viainfluence(fitted_model)) (John Fox) (GH #672)
predict works for formulas containing . whennewdata is specified (GH #653)
bootMer now correctly inherits control settings from original fit

CHANGES IN VERSION 1.1-28 (2022-02-04)

USER-VISIBLE CHANGES

construction of interacting factors (e.g. whenf1:f2 orf1/f2 occur in random effects terms) is now more efficientfor partially crossed designs(doesn't try to create all combinations off1 andf2) (GH #635 and #636)
mkNewReTrms is exported
singular-fit message now refers tohelp("isSingular")rather than?isSingular

TESTS

fixall.equal(p1,p2,p3) and similarexpect_equal() thinkos
fix some tests only run whenlme4:::testLevel() > 1;adapt tests for upcoming Matrix 1.4-1 which hasnames(diag(<sparse>))

BUG FIXES

reOnly preserves environment (GH #654, Mikael Jagan)
backward-compatibility hooks changed to evaluate at run-time(i.e., in.onLoad()) rather than at build time (GH #649)
lmList no longer warns whendata is a tibble(GH #645)

CHANGES IN VERSION 1.1-27.1 (2021-06-22)

USER-VISIBLE CHANGES

influence.merMod allows user-specified starting parameters
cleaned up performance vignette

BUG FIXES

cooks.distance now works with objects computedbyinfluence method
influence.merMod now works withglmer modelsusingnAGQ=0
predict (with new data) andsimulate methodsnow work for models with >100 levels in a random effect groupingvariable (GH #631)

CHANGES IN VERSION 1.1-27 (2021-05-15)

USER-VISIBLE CHANGES

improvements from Lionel Henry (viahttps://github.com/lme4/lme4/pull/587) to fix corner cases indata checking; also resolves GH #601 (allFit scoping)
getME(., "lower") now has names (request of GH #609)
improved detection ofNaN in internal calculations(typically due to underflow/overflow or out-of-bounds linearpredictors from non-constraining link functions such as identity-linkGamma models)
influence.merMod allows parallel computation
thestatmod package is no longer required unless attemptingto simulate results from a model with an inverse Gaussian response

BUG FIXES

long formulas work better inanova headings (GH #611)

CHANGES IN VERSION 1.1-26 (2020-11-30)

BUG FIXES

predict,model.frame(.,fixed.only=TRUE) workwith variable names containing spaces (GH #605)
simulate works when original response variable waslogical
densityplot handles partly broken profiles more robustly

NEW FEATURES

thpr method fordensityplot() (for plottingprofiles scaled as densities) gets new arguments

CHANGES IN VERSION 1.1-25 (2020-10-23)

Set more tests to run only if environment variableLME4_TEST_LEVEL>1

CHANGES IN VERSION 1.1-24 (never on CRAN)

USER-VISIBLE CHANGES

anova() now returns a p-value ofNA if the dfdifference between two models is 0 (implying they are equivalentmodels) (GH#583, @MetaEntropy)
speedup incoef() for large models, by skippingconditional variance calculation (Alexander Bauer)
simulate.formula machinery has changed slightly,for compatibility with theergm package (Pavel Krivitsky)
informational messages about (non-)convergence improved (GH#599)
improved error messages for 0 non-NA cases in data (GH #533)

NEW FEATURES

getME(.,"devfun") now works forglmer objects.Additionally,profile/confint for GLMMsno longer depend on objects in the fitting environmentremaining unchanged (GH #589). This change also affects likelihoodprofiling machinery; results ofglmer profiling/CIs maynot match results from previous versions exactly.

BUG FIXES

improved handling/documentation ofglmer.nb controls(GH #556)
predict works better forgamm4 objects (GH#575)
resolved some long-standing UBSAN issues (GH #561)

CHANGES IN VERSION 1.1-23 (2020-03-06)

This is primarily for CRAN compliance (previous submission wasretracted to allow time for downstream package adjustments).

Some PROTECT/UNPROTECT fixes

CHANGES IN VERSION 1.1-22 (never on CRAN)

USER-VISIBLE CHANGES

prediction now works better for factors with many levels(GH#467, solution by @sihoward)
minor changes to argument order in[g]lmerControl;default tolerance for convergence checks increased from 0.001 to 0.002forglmerControl (now consistent withlmerControl)
lmer(*, family="<fam>") is no longer valid; it had beendeprecated since 2013-06.
lmer(),glmer(), andnlmer() no longerhave a formal... argument. This defunctifies the use of asparseX = . argument and will reveal some user errors, whereextraneous arguments were previously disregarded.
InisSingular(x, tol), the default tolerance (tol) hasbeen increased from1e-5 to1e-4, the default ofcheck.conv.singular ing?lmerControl().
for clarity and consistency with base R methods,some column names ofanova() output are changed:"Df" becomes "npar", "Chi Df" becomes "Df" (GH #528)
simulate() now works with inverse-Gaussian models(GH #284 revisited, @nahorp/Florian Hartig)
single-model mode ofanova()now warns about unused arguments in ...(e.g.type="III")
default tolerances fornloptwrap/BOBYQA optimizertightened (xtol_abs andftol_abs were 1e-6,now 1e-8). (To revert to former tolerances, usecontrol=lmerControl(optimizer="nloptwrap",optCtrl=list(xtol_abs=1e-6, ftol_abs=1e-6)).)

BUG FIXES

improved checking for missing data (@lionel-)
internalcheckZrank() should be able to deal with(Matrix package)rankMatrix() returningNA.
allFit(fm) now works for a model that had an explicitcontrol = lmerControl(..) call.
internalgetStart() now works when model'sstart was specified as a list,and when called fromdrop1() ona submodel, fixing GH #521.
internal functionmkdevfun now works even if there isan extraneousgetCall function defined in the globalenvironment (GH #535)
allFit() works even if a variable with symboli is used somewhere in the original model call (GH #538,reported by Don Cohen); generally more robust
glmer.nb works even if an alternative versionofnegative.binomial (other than the one fromMASS)is loaded in the workspace(e.g. by theGLMMadaptive package) (GH#516)
level argument is now honoured byconfint(..., type="boot", level=...) (GH #543)

CHANGES IN VERSION 1.1-21 (2019-03-05)

USER-VISIBLE CHANGES

bootMer now traps and stores messages, warnings, and errors
bootMer returns an object of classc("bootMer","boot"); newprint andconfint methods for classbootMer
small changes to wording of singular-fit messages

CHANGES IN VERSION 1.1-20 (2019-02-04)

USER-VISIBLE CHANGES

default value forcondVar (whether to return conditional variancesas part of theranef.merMod object) is nowTRUE
changed default optimizer to "nloptwrap" (BOBYQAimplementation from thenloptr package) forlmermodels. To revert to the old default, usecontrol=lmerControl(optimizer="bobyqa")

BUG FIXES

adapted tests to work with R-devel's more consistentformula(model.frame(.)) behavior.

CHANGES IN VERSION 1.1-19 (2018-11-10)

NEW FEATURES

influence measure code fromcar rolled in (see?influence.merMod)
mkReTrm gets new argumentsreorder.terms,reorder.vars to control arrangement of RE terms andindividual effects with RE terms within model structures
adding material from theRePsychLing package(on GitHub; see Bates et al 2015 arXiv:1506.04967) toshow orthogonal variance components.
new utilityisSingular() function fordetecting singular fits
allFit function/methods have been moved to the mainpackage, rather than being included in an auxiliary source file;computations can (in principle) be done in parallel

USER-VISIBLE CHANGES

by default a message is now printed for singular fits(i.e., fits with linear combinations of variance componentsthat are exactly zero)
as.data.frame.merMod finds conditional varianceinformation stored either asattr(.,"postVar") orattr(.,"condVar") (forglmmTMB compatibility)
change to defaults of[g]lmerControl to print amessage when fits are singular
post-fitting convergence checks based on estimated gradientand Hessian (seetroubleshooting) are no longerperformed for (nearly-)singular fits (seeisSingular)

CHANGES IN VERSION 1.1-18-1 (2018-08-17)

This is a minor release; the only change is to roll back (unexport) theinfluence.merMod method, pending resolution of conflicts with the car package

CHANGES IN VERSION 1.1-18 ((2018-08-16)

USER-VISIBLE CHANGES

ranef(.,condVar=TRUE) now works when thereare multiple random effects terms per factor

NEW FEATURES

rstudent andinfluence methods are availableformerMod objects
devfun2 function (for generating a deviancefunction that works on the standard deviation/correlation scale)is now exported

BUG FIXES

lmList now obeys itspool argument (instead ofalways using what currently is the default, GH #476)

CHANGES IN VERSION 1.1-17 (2018-04-03)

This is a maintenance release only (fixes CRAN problems withcross-platform tests and examples)

CHANGES IN VERSION 1.1-16 (2018-03-28)

BUG FIXES

lmList no longer ignores thesubset argument(John Fox)
fixed several minor issues with predicting when (1) groupingvariables have different levels from original model (e.g. missinglevels/factor levels not explicitly specified innewdata) or (2)re.form is a subset of the original RE formula and some(unused) grouping variables are omitted fromnewdata (GH#452, #457)

USER-VISIBLE CHANGES

lmList tries harder to collect errors and pass themon as warnings
documentedas.function method (given amerModobject, returns a function that computes the deviance/REML criterion forspecified parameters)
print method forsummary.merMod objectsno longer collapses small values of the t-statistic to zero

CHANGES IN VERSION 1.1-15 (2017-12-21)

BUG FIXES

model.frame(., fixed.only=TRUE) now handles models with"non-syntactic" (e.g. space-containing/backtick-delimited)variables in the formula.
confint(<merMod>) now works again for the defaultmethod"profile".

USER-VISIBLE CHANGES

exporteddotplot.ranef.mer

CHANGES IN VERSION 1.1-14 (2017-09-27)

Primarily an R-devel/CRAN-compatibility release.

NEW FEATURES

addedtransf argument todotplot.ranef.merto allow back-transformation (Ferenci Tamás, GH #134)
addedas.data.frame.ranef.mer convenience method
user can specify initial value for overdispersion parameteringlmer.nb (Timothy Lau, GH #423)

BUG FIXES

fix bug where NAs in fitting data were carried over intopredictions on new data (!) (lmwang9527, GH #420)
fix bug with long terms in models with || notation
nlmer now respects user-specified lower/upper bounds(GH #432)
confint.thpr (confint method applied to analready-computed profile now respects "theta_"/"beta_"specifications to return all random-effect or all fixed-effectconfidence intervals, respectively.

DOCUMENTATION IMPROVEMENTS

document need to export packages and objects to workerswhen usingbootMer with snow

USER-VISIBLE CHANGES

improved warning message when usinglmerControl() withglmer (GH #415)
avoid deparsing big data frames when checking data (GH#410)
passverbose options to nloptr optimizers whenusingnloptwrap (previously ignored, with a warning)
thefl (factor list) component ofmkReTrmsobjects is now returned as a list rather than a data frame

CHANGES IN VERSION 1.1-13 (2017-04-18)

NEW FEATURES

addedprof.scale argument toprofile.merMod,documented caveats about usingvarianceProf/logProftransformation methods for correlation parameters

BUG FIXES

suppressed spurious contrast-dropping warning (GH #414)
fixed bug inconfint.lmList4 (GH #26)
fixed bug whenFUN returned an unnamedvector inconfint(.,FUN=FUN,method="boot")
fixed small bug relating tonAGQ0initStep=FALSE

CRAN-COMPATIBILITY UPDATES

fixed time stamps on compiled versions of vignettes

CHANGES IN VERSION 1.1-12 (2016-04-15)

This release is primarily a bump for compatibility with the newWindows toolchain. Some small documentation and test changes.

USER-VISIBLE CHANGES

reduced default print precision of fixed-effect correlationmatrix insummary.merMod (related to GH #300)

BUG FIXES

fixed bug inde novo Gamma-response simulations

CHANGES IN VERSION 1.1-11 (2016-02-11)

USER-VISIBLE CHANGES

changeVarCorr method signature (for compatibilitywith upstreamnlme changes)

BUG FIXES

severalglmer.nb bugs fixed (generally notchanging results, but causing warnings and errors e.g.during bootstrapping)
fixes to somelmList bugs (Github #320)
minor documentation, vignette updates
minor fix toplot.merMod withid specified
bootMer now handles separate offset term properly(Github #250)

CHANGES IN VERSION 1.1-10 (2015-10-05)

This release is primarily a version bump for the release ofthe paper in J. Stat. Software.

USER-VISIBLE CHANGES

updated CITATION file.

NEW FEATURES

We export set of about a dozen printing utility functionswhich are used in ourprint methods.
bootMer now allows the use ofre.form.

BUG FIXES

fixed reordering bug in names ofgetME(.,"Ztlist")(terms are reordered in decreasing order of the number oflevels of the grouping variable, but names were not beingreordered)
fixed issue with simulation when complex forms (suchas nested random effects terms) are included in the model(Github #335)

CHANGES IN VERSION 1.1-9 (2015-08-20)

USER-VISIBLE CHANGES

explicitmaxit arguments for various functions(refit,mkGlmerDevfun, ...)

NEW FEATURES

terms andformula methods now haverandom.only options
getME gains aglmer.nb.theta option.It is now (an S3) generic with an"merMod" method inlme4 and potentially other methods in dependent packages.
simulate now works forglmer.nb models(Github #284: idea from @aosmith16)

BUG FIXES

prediction and simulation now work when random-effectsterms have data-dependent bases (e.g.,poly(.) orns(.) terms) (Github #313, Edgar Gonzalez)
logLik forglmer.nb models nowincludes the overdispersion parameter in theparameter count (df attribute)
lmList handles offsets and weights better
lots of fixes toglmer.nb (Github #176, #266, #287,#318).Please note that glmer.nb is still somewhatunstable/under construction.

CRAN-COMPATIBILITY UPDATES

import functions from base packages to pass CRAN checks
tweak to failing tests on Windows

CHANGES IN VERSION 1.1-8 (2015-06-22)

NEW FEATURES

getME gains a"Tlist" option(returns a vector of template matrices from which the blocks ofLambda are generated)
hatvalues method returns the diagonal of the hatmatrix of LMMs
nlminbwrap convenience function allows use ofnlminb without going through theoptimx package
as.data.frame.VarCorr.merMod gains anorderoption that allows the results to be sorted with variances firstand covariances last (default) or in lower-triangle order
allow more flexibility inscales forxyplot.thpr method (John Maindonald)
models with only random effects of the form1|f havebetter starting values forlmer optimization (GaborGrothendieck)
glmer now allows a logical vector as the responsefor binomial models
anova will now do (sequential) likelihood ratiotests for two or more models including bothmerMod andglm orlm models (at present, only for GLMMsfitted with the Laplace approximation)

USER-VISIBLE CHANGES

deviance() now returns the deviance, rather than half thenegative log-likelihood, for GLMMs fitted with Laplace(the behaviour for LMMs and GLMMs fitted withnAGQ>1 hasnot changed)
convergence warning and diagnostic test issues are nowreported inprint andsummary methods
update now (attempts to) re-evaluate the original fitin the environment of its formula (as is done withdrop1)
refit of a nonlinear mixed model fit now throws anerror, but this will hopefully change in future releases (relatedto bug fixes for Github #231)
lmList now returns objects of classlmList4,to avoid overwritinglmList methods from the recommendednlme package
names of random effects parameters inconfintchanged (modified for consistency across methods);oldNames=TRUE (default) gives".sig01"-style names,oldNames=FALSE gives"sd_(Intercept)|Subject"-style names
confint(.,method="Wald") result now contains rowsfor random effects parameters (values set toNA)as well as for fixed-effect parameters

BUG FIXES

simulate andpredict now work moreconsistently with different-length data, differingfactor levels, andNA values (Github #153, #197, #246, #275)
refit now works correctly forglmerfits (Github #231)
fixed bug infamily.merMod; non-default linkswere not retrieved correctly (Alessandro Moscatelli)
fixedbootMer bug fortype=="parametric",use.u=TRUE (Mark Lai)
gradient scaling for convergence checks now uses the Cholesky factor of theHessian; while it is more correct, this will lead to some additional (probablyfalse-positive) convergence warnings
As withlm(), users now get an error for non-finite(Inf,NA, orNaN)values in the response unlessna.action is set toexclude or omit them (Github #310)

CHANGES IN VERSION 1.1-7 (2014-07-19)

NEW FEATURES

thenloptr package is now imported;a wrapper function (nloptwrap) is provided so thatlmerControl(optimizer="nloptwrap") is all that's necessaryto usenloptr optimizers in the nonlinear optimizationstage (the default algorithm is NLopt's implementationof BOBYQA: see?nloptwrap for examples)
preliminary implementation of checks for scalingof model matrix columns (seecheck.scaleX in?lmerControl)
beta is now allowed as a synonym forfixefwhen specifying starting parameters (Github #194)

USER-VISIBLE CHANGES

the use ofdeviance to return the REML criterionis now deprecated; users should useREMLcrit() instead(Github #211)
changed the default value ofcheck.nobs.vs.rankZ to"ignore" (Github #214)

BUG FIXES

change gradient testing from absolute to relative
fixconfint(.,method="boot") to allow/workproperly withboot.type values other than"perc"(reported by Alan Zaslavsky)
allowplot() to work when data are specified in a differentenvironment (reported by Dieter Menne)
predict andsimulate work for matrix-valuedpredictors (Github #201)
othersimulate bugs (Github #212)
predict no longer warns spuriously whenoriginal response was a factor (Github #205)
fix memory access issues (Github #200)

CHANGES IN VERSION 1.1-6 (2014-04-13)

This version incorporates no changes in functionality, justmodifications to testing and dependencies for CRAN/backward compatibility.

BUG FIXES

changedrop1 example to prevent use of old/incompatiblepbkrtest versions, for 2.15.3 compatibility
explicitlyrequire(mlmRev) for tests to prevent cyclicdependency
bumpRcppEigen Imports: requirement from >0.3.1.2.3 to>=0.3.2.0;Rcpp dependency to >= 0.10.5

CHANGES IN VERSION 1.1-5 (2014-03-14)

BUG FIXES

improved NA handling insimulate andrefit
made internal handling ofweights/offsetarguments slightly more robust (Github #191)
handle non-positive-definite estimated fixed effectvariance-covariance matrices slightly more generally/robustly(fall back on RX approximation, with a warning,if finite-difference Hessianis non-PD; returnNA matrix if RX approximation isalso bad)

MINOR USER-VISIBLE CHANGES

Added output specifying when Gauss-Hermite quadraturewas used to fit the model, and specifying number of GHQ points(Github #190)

CHANGES IN VERSION 1.1-4

BUG FIXES

Models with prior weights returned an incorrect sigma anddeviance (Github issue #155). The deviance bug was only apractical issue in model comparisons, not with inferences given aparticular model. Both bugs are now fixed.
Profiling failed in some cases for models with vector randomeffects (Github issue #172)
Standard errors of fixed effects are now computedfrom the approximate Hessian by default (see theuse.hessian argument invcov.merMod); thisgives better (correct) answers when the estimates ofthe random- and fixed-effect parameters are correlated(Github #47)

MAJOR USER-VISIBLE CHANGES

The default optimizer forlmer fits has beenswitched from "Nelder_Mead" to "bobyqa" because we havegenerally found the latter to be more reliable. To switchback to the old behaviour,usecontrol=lmerControl(optimizer="Nelder_Mead").
Better handling of rank-deficient/overparameterizedfixed-effect model matrices; seecheck.rankX optionto[g]lmerControl. The default value is"message+drop.cols", which automatically drops redundantcolumns and issues a message (not a warning). (Github #144)

MINOR USER-VISIBLE CHANGES

slight changes in convergence checking; tolerances canbe specified where appropriate, and some default toleranceshave changed (e.g.,check.conv.grad)
improved warning messages about rank-deficiency in X and Zetc. (warnings now try to indicate whether the unidentifiabilityis in the fixed- or random-effects part of the model)
predict andsimulate now preferre.form as the argument to specify which random effectsto condition on, but allowReForm,REForm, orREform, giving a message (not a warning) that they aredeprecated (addresses Github #170)
small fixes for printing consistency in models with nofixed effects
we previously exported afortify function identicalto the one found inggplot2 in order to be able to define afortify.merMod S3 method without inducing a dependency onggplot2. This has now been unexported to avoid maskingggplot2's ownfortify methods; if you want toadd diagnostic information to the results of a model, usefortify.merMod explicitly.
simulate.formula now checks for names associatedwith thetheta andbeta parameter vectors. Ifmissing, it prints a message (not a warning); otherwise, itre-orders the parameter vectors to match the internalrepresentation.
preliminary implementation of acheck.scaleX argumentin[g]lmerControl that warns about scaling if some columnsof the fixed-effect model matrix have large standarddeviations (relative to 1, or to each other)

CHANGES IN VERSION 1.1-3

NEW FEATURES

The gradient and Hessian are now computed via finitedifferencing after the nonlinear fit is done, and the resultsare used for additional convergence tests. Control of thebehaviour is available through thecheck.conv.* optionsin[g]lmerControl. Singular fits (fits with estimatedvariances of zero or correlations of +/- 1) can also be tested for,although the current default value of thecheck.conv.singularoption is"ignore"; this may be changed to"warning"in the future. The results are stored in@optinfo$derivs.(Github issue #120; based on code by Rune Christensen.)
Thesimulate method will now work to generatesimulations "from scratch" by providing a model formula,a data frame holding the predictor variables, and a listcontaining the values of the model parameters:see?simulate.merMod. (Github issue #115)
VarCorr.merMod objects now have anas.data.framemethod, converting the list of matrices to a moreconvenient form for reporting and post-processing. (Github issue #129)

MINOR USER-VISIBLE CHANGES

results offitted(),predict(),andresiduals() now havenames in all cases (previously results were unnamed, ornamed only when predicting from new data)
theanova method now has arefit argumentthat controls whether objects of classlmerMod should berefitted with ML before producing theanova table.(Github issues #141, #165; contributed by Henrik Singmann.)
theprint method forVarCorr objectsnow has aformatter argument for finer controlof standard deviation and variance formats
theoptinfo slot now stores slightly moreinformation, including the number of function evaluations($feval).
dotplot.ranef.mer now adds titles to sub-plots by default,likeqqmath.ranef.mer

BUG FIXES

fitted now respectsna.action settings (Githubissue #149)
confint(.,method="boot") now works when there areNA values in the original data set (Github issue #158)
previously, the code stored the results (parameter values,residuals, etc.) based on thelast set of parametersevaluated, rather than the optimal parameters. These werenot always the same, but were almost always very close,but some previous results will change slightly(Github issue #166)

CHANGES IN VERSION 1.1-0

MINOR USER-VISIBLE CHANGES

when using the defaultmethod="profile",confint now returns appropriate upper/lower bounds(-1/1 for correlations, 0/Inf for standard deviations)rather thanNA when appropriate

BUG FIXES

in a previous development version,ranef returnedincorrect conditional variances (github issue #148). this isnow fixed

CHANGES IN VERSION 1.0-6 (2014-02-02)

BUG FIXES

prediction now works when new data have fewer factorlevels than are present in the original data (Github issue #143,reported by Rune Haubo)
the existence of a variable "new" in the global environmentwould messlme4 up: reported athttp://stackoverflow.com/questions/19801070/error-message-glmer-using-r-what-must-be-a-character-string-or-a-function

CHANGES IN VERSION 1.0-5 (2013-10-24)

USER-VISIBLE CHANGES

confint.merMod andvcov.merMod arenow exported, for downstream package-author convenience
the package now depends on Matrix >=1.1-0 and RcppEigen>=0.3.1.2.3
newrename.response option forrefit (see BUGFIXES section)

BUG FIXES

eliminated redundant messages about suppressedfixed-effect correlation matrices when p>20
most inverse-link functions are now bounded whereappropriate by.Machine$double.eps, allowing fittingof GLMMs with extreme parameter values
merMod objects created withrefit did notwork withupdate: optionalrename.response option added torefit.merMod, to allowthis (but the default is stillFALSE, forback-compatibility) (reported by A. Kuznetsova)
fixed buglet preventing on-the-fly creation of index variables,e.g.y~1+(1|rownames(data)) (reported by J. Dushoff)
predict now works properly forglmer modelswith basis-creating terms (e.g.poly,ns)
step sizes determined from fixed effect coefficient standarderrors after first state ofglmer fitting are now bounded,allowing some additional models to be fitted

CHANGES IN VERSION 1.0-4 (2013-09-08)

BUG FIXES

refit() now works, again, with lists oflength 1, so that e.g.refit(.,simulate(.)) works.(Reported by Gustaf Granath)
getME(.,"ST") was returning a listcontaining the Cholesky factorizations that get repeated inLambda. But this was inconsistent with whatST represents inlme4.0. This inconsistency has now been fixed andgetME(.,"ST") is now consistent with the definition of theST matrix inlme4.0. Seehttps://github.com/lme4/lme4/issues/111 for moredetail. Thanks to Vince Dorie.
Corrected order of unpacking of standarddeviation/correlation components, which affected resultsfromconfint(.,method="boot"). (Reported by ReinholdKliegl)
fixed a copying bug that maderefitML()modify the original model

CHANGES IN VERSION 1.0-1 (2013-08-17)

MINOR USER-VISIBLE CHANGES

check.numobs.* andcheck.numlev.* in(g)lmerControl have been changed (from recent developmentversions) tocheck.nobs.* andcheck.nlev.* respectively, and the default values ofcheck.nlev.gtreq.5 andcheck.nobs.vs.rankZhave been changed to"ignore" and"warningSmall"respectively
in(g)lmerControl, arguments to the optimizershould be passed as a list calledoptCtrl, rather thanspecified as additional (ungrouped) arguments
thepostVar argument toranef has beenchanged to the (more sensible)condVar ("posterior variance"was a misnomer, "conditional variance" – short for "variance of theconditional mode" – is preferred)
theREform argument topredict has been changedtoReForm for consistency
thetnames function, briefly exported, has beenunexported
getME(.,"cnms") added
print method formerMod objects is now moreterse, and different fromsummary.merMod
theobjective method for therespModreference class now takes an optionalsigma.sq parameter(defaulting toNULL) to allow calculation of theobjective function with a residual variance different fromthe profiled value (Vince Dorie)

CHANGES IN VERSION 1.0-0 (2013-08-01)

MAJOR USER-VISIBLE CHANGES

Because the internal computational machinery has changed,results from the newest version oflme4 will not be numericallyidentical to those from previous versions. For reasonably well-defined fits, they will be extremely close (within numericaltolerances of 1e-4 or so), but for unstable or poorly-defined fitsthe results may change, and very unstable fits may fail when they(apparently) succeeded with previous versions. Similarly, some fitsmay be slower with the new version, although on average the newversion should be faster and more stable. More numericaltuning options are now available (see below); non-default settingsmay restore the speed and/or ability to fit a particular model withoutan error. If you notice significant or disturbing changes when fittinga model with the new version oflme4,please notify the maintainers.
VarCorr returns its results in the same format as before (as alist of variance-covariance matrices withcorrelation andstddevattributes, plus asc attribute giving the residual standarddeviation/scale parameter when appropriate), but prints them in adifferent (nicer) way.
By defaultresiduals gives deviance (rather than Pearson)residuals when applied toglmer fits (a side effect of matchingglmbehaviour more closely).
As another side effect of matchingglmbehaviour, reported log-likelihoods fromglmer modelsare no longer consistent with those from pre-1.0lme4,butare consistent withglm; seeglmerexamples.

MINOR USER-VISIBLE CHANGES

More use is made of S3 rather than S4 classes and methods: oneside effect is that thenlme andlme4 packages are now much morecompatible; methods such asfixef no longer conflict.
The internal optimizer has changed.[gn]lmer now has anoptimizer argument;"Nelder_Mead" is the default for[n]lmer,while a combination of"bobyqa" (an alternative derivative-freemethod) and"Nelder_Mead" is the default forglmer. To use thenlminb optimizer as in the old version oflme4, you can useoptimizer="optimx" withcontrol=list(method="nlminb") (you willneed theoptimx package to be installed and loaded). SeelmerControl for details.
Families in GLMMs are no longer restricted to built-in/hard-coded families; any family described infamily, or following thatdesign, is usable (although there are some hard-coded families, whichwill be faster).
[gn]lmer now produces objects of classmerMod rather thanclassmer as before.
the structure of theZt (transposed random effectdesign matrix) as returned bygetME(.,"Zt"), and thecorresponding order of the random effects vector(getME(.,"u")) have changed. To retrieveZtin the old format, usedo.call(Matrix::rBind,getME(.,"Ztlist")).
the package checks input more thoroughly fornon-identifiable or otherwise problematic cases: seelmerControl for fine control of the test behaviour.

NEW FEATURES

A general-purposegetME accessor method allowsextraction of a wide variety of components of a mixed-modelfit.getME also allows a vector of objects to be returned asa list of mixed-model components. This has been backported tobe compatible with older versions oflme4 that still producemerobjects rather thanmerMod objects. However, backporting is incomplete;some objects are only extractable in newer versions oflme4.
Optimization information (convergence codes, warnings, etc.)is now stored in an@optinfo slot.
bootMer provides a framework for obtaining parameter confidenceintervals by parametric bootstrapping.
plot.merMod provides diagnostic plottingmethods similar to those from thenlme package(although missingaugPred).
Apredict.merMod method gives predictions;it allows an effect-specific choice of conditional prediction or prediction at thepopulation level (i.e., with random effects set to zero).
Likelihood profiling forlmer andglmer results (seelink{profile-methods}).
Confidence intervals by likelihood profiling (default),parametric bootstrap, or Wald approximation (fixed effects only):seeconfint.merMod
nAGQ=0, an option to do fast (but inaccurate) fitting of GLMMs.
UsingdevFunOnly=TRUE allows the user to extract a deviancefunction for the model, allowing further diagnostics/customization ofmodel results.
The internal structure of [gn]lmer is now more modular, allowingfiner control of the different steps of argument checking; constructionof design matrices and data structures; parameter estimation; and constructionof the finalmerMod object (see?modular).
theformula,model.frame, andtermsmethods return full versions (including random effect terms andinput variables) by default, but afixed.only argumentallows access to the fixed effect submodel.

EXPERIMENTAL FEATURES

glmer.nb provides an embryonic negativebinomial fitting capability.

STILL NON-EXISTENT FEATURES

Adaptive Gaussian quadrature (AGQ) is not available for multiple and/ornon-scalar random effects.
Posterior variances of conditional models for non-scalar random effects.
Standard errors forpredict.merMod results.
Automatic MCMC sampling based on the fit turns out to be very difficultto implement in a way that is really broadly reliable and robust;mcmcsampwill not be implemented in the near future. Seepvalues for alternatives.
"R-side" structures (within-block correlation and heteroscedasticity) arenot on the current timetable.

BUG FIXES

In a development version, prior weights were not being used properly inthe calculation of the residual standard deviation, but this has been fixed.Thanks to Simon Wood for pointing this out.
In a development version, the step-halving component of the penalizediteratively reweighted least squares algorithm was not working, butthis is now fixed.
In a development version, squareRZX matrices would lead to apwrssUpdate did not converge in 30 iterations error. This has been fixedby adding an extra column of zeros toRZX.

DEPRECATED AND DEFUNCT

Previous versions oflme4 providedthemcmcsamp function, which efficiently generateda Markov chain Monte Carlo sample from the posteriordistribution of the parameters, assuming flat (scaledlikelihood) priors. Due to difficulty in constructing aversion ofmcmcsamp that was reliable even incases where the estimated random effect variances werenear zero (e.g.https://stat.ethz.ch/pipermail/r-sig-mixed-models/2009q4/003115.html),mcmcsamp has been withdrawn (or more precisely,not updated to work withlme4 versions >=1.0).
Callingglmer with the defaultgaussian familyredirects tolmer, but this is deprecated(in the futureglmer(...,family="gaussian") mayfit a LMM using the penalized iteratively reweighted least squaresalgorithm). Please calllmer directly.
Callinglmer with afamily argument redirectstoglmer; this is deprecated. Please callglmer directly.

CHANGES IN VERSION 0.999375-16 (2008-06-23)

MAJOR USER-VISIBLE CHANGES

The underlying algorithms and representations for all themixed-effects models fit by this package have changed - forthe better, we hope. The class "mer" is a commonmixed-effects model representation for linear, generalizedlinear, nonlinear and generalized nonlinear mixed-effectsmodels.
ECME iterations are no longer used at all, nor are analyticgradients. Components named 'niterEM', 'EMverbose', or'gradient' can be included in the 'control' argument tolmer(), glmer() or nlmer() but have no effect.
PQL iterations are no longer used in glmer() and nlmer().Only the Laplace approximation is currently available. AGQ,for certain classes of GLMMs or NLMMs, is being added.
The 'method' argument to lmer(), glmer() or nlmer() isdeprecated. Use the 'REML = FALSE' in lmer() to obtain MLestimates. Selection of AGQ in glmer() and nlmer() will becontrolled by the argument 'nAGQ', when completed.

NEW FEATURES

The representation of mixed-effects models has beendramatically changed to allow for smooth evaluation of theobjective as the variance-covariance matrices for the randomeffects approach singularity. Beta testers found thisrepresentation to be more robust and usually faster thanprevious versions of lme4.
The mcmcsamp function uses a new sampling method for thevariance-covariance parameters that allows recovery fromsingularity. The update is not based on a sample from theWishart distribution. It uses a redundant parameterrepresentation and a linear least squares update.
CAUTION: Currently the results from mcmcsamp look peculiar andare probably incorrect. I hope it is just a matter of myomitting a scaling factor but I have seen patterns such asthe parameter estimate for some variance-covariance parametersbeing the maximum value in the chain, which is highlyunlikely.
The 'verbose' argument to lmer(), glmer() and nlmer() can beused instead of 'control = list(msVerbose = TRUE)'.