[Lme4-commits] r1743 - www/misc
noreply at r-forge.r-project.org
noreply at r-forge.r-project.org
Wed May 16 23:17:43 CEST 2012
Author: bbolker
Date: 2012-05-16 23:17:42 +0200 (Wed, 16 May 2012)
New Revision: 1743
Modified:
www/misc/lme4_compat_report.html
www/misc/lme4_conversion.html
www/misc/lme4_conversion.md
Log:
updated conversion / compatibility info
Modified: www/misc/lme4_compat_report.html
===================================================================
--- www/misc/lme4_compat_report.html 2012-05-16 20:16:32 UTC (rev 1742)
+++ www/misc/lme4_compat_report.html 2012-05-16 21:17:42 UTC (rev 1743)
@@ -22,7 +22,7 @@
<p class='character'>lme4 version: 0.999902344.0</p>
-<p class='character'>Test date: Tue May 15 13:46:03 2012</p>
+<p class='character'>Test date: Wed May 16 15:02:30 2012</p>
<h2 > Notes</h2>
@@ -118,7 +118,7 @@
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
-<td class=cellinside>63: could not find function "GHrule"<br>64: Calls: iccbin ... withCallingHandlers -> do.call -> <Anonymous> -> .Call -> cpp_exception<br>65: Execution halted
+<td class=cellinside>62: Error in nrow(GQmat) : could not find function "GHrule"<br>63: Calls: iccbin ... do.call -> <Anonymous> -> fn1 -> fn -> pwrssUpdate -> nrow<br>64: Execution halted
</td></tr>
<tr>
@@ -192,13 +192,13 @@
</td>
<td class=cellinside>CRAN
</td>
-<td class=cellinside>0.9-6
+<td class=cellinside>1.0-0
</td>
<td class=cellinside>Heather Turner
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
-<td class=cellinside>74: object 'expandSlash' not found<br>75: Calls: BTm -> do.call -> <Anonymous> -> Diff -> ::: -> get<br>76: Execution halted
+<td class=cellinside>67: Error in match.arg(type) : 'arg' should be one of "link", "response"<br>68: Calls: predict -> predict.glmmPQL -> match.arg<br>69: Execution halted
</td></tr>
<tr>
@@ -358,7 +358,7 @@
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
-<td class=cellinside>112: cannot coerce type 'closure' to vector of type 'character'<br>113: Calls: ezBoot ... formula -> formula.default -> eval -> eval -> eval -> parse<br>114: Execution halted
+<td class=cellinside>114: cannot coerce type 'closure' to vector of type 'character'<br>115: Calls: ezBoot ... formula -> formula.default -> eval -> eval -> eval -> parse<br>116: Execution halted
</td></tr>
<tr>
@@ -390,7 +390,7 @@
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
-<td class=cellinside>74: Error in b$FL : $ operator not defined for this S4 class<br>75: Calls: gamm4<br>76: Execution halted
+<td class=cellinside>76: Error in b$FL : $ operator not defined for this S4 class<br>77: Calls: gamm4<br>78: Execution halted
</td></tr>
<tr>
@@ -460,38 +460,6 @@
<tr>
<td class=firstcolumn>26
</td>
-<td class=cellinside>HSAUR
-</td>
-<td class=cellinside>CRAN
-</td>
-<td class=cellinside>1.3-0
-</td>
-<td class=cellinside>Torsten Hothorn
-</td>
-<td class=cellinside><font style="color:red">error_vignette</font>
-</td>
-<td class=cellinside>4857: Error: 'arg' should be one of "Laplace", "AGQ"<br>4858: Execution halted<br>4859:
-</td></tr>
-
-<tr>
-<td class=firstcolumn>27
-</td>
-<td class=cellinside>HSAUR2
-</td>
-<td class=cellinside>CRAN
-</td>
-<td class=cellinside>1.1-1
-</td>
-<td class=cellinside>Torsten Hothorn
-</td>
-<td class=cellinside><font style="color:red">error_vignette</font>
-</td>
-<td class=cellinside>107: Error: PIRLS step failed<br>108: Execution halted<br>109:
-</td></tr>
-
-<tr>
-<td class=firstcolumn>28
-</td>
<td class=cellinside>influence.ME
</td>
<td class=cellinside>CRAN
@@ -506,7 +474,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>29
+<td class=firstcolumn>27
</td>
<td class=cellinside>kulife
</td>
@@ -522,7 +490,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>30
+<td class=firstcolumn>28
</td>
<td class=cellinside>languageR
</td>
@@ -538,7 +506,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>31
+<td class=firstcolumn>29
</td>
<td class=cellinside>LMERConvenienceFunctions
</td>
@@ -546,7 +514,7 @@
</td>
<td class=cellinside>1.6.8.2
</td>
-<td class=cellinside>Antoine Tremblay
+<td class=cellinside>Antoine Tremblay; IWK Health Center
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
@@ -554,7 +522,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>32
+<td class=firstcolumn>30
</td>
<td class=cellinside>longpower
</td>
@@ -570,7 +538,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>33
+<td class=firstcolumn>31
</td>
<td class=cellinside>MEMSS
</td>
@@ -586,7 +554,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>34
+<td class=firstcolumn>32
</td>
<td class=cellinside>MethComp
</td>
@@ -596,30 +564,14 @@
</td>
<td class=cellinside>Bendix Carstensen
</td>
-<td class=cellinside><font style="color:purple">error_depfail</font>
+<td class=cellinside><font style="color:blue">OK</font>
</td>
-<td class=cellinside>9: Package suggested but not available: 'BRugs'
+<td class=cellinside>
</td></tr>
<tr>
-<td class=firstcolumn>35
+<td class=firstcolumn>33
</td>
-<td class=cellinside>mi
-</td>
-<td class=cellinside>CRAN
-</td>
-<td class=cellinside>0.09-16
-</td>
-<td class=cellinside>Yu-Sung Su
-</td>
-<td class=cellinside><font style="color:red">error_install</font>
-</td>
-<td class=cellinside>** data<br>** inst<br>** preparing package for lazy loading<br>Error : objects 'coef', 'fitted', 'resid', 'residuals', 'plot' are not exported by 'namespace:lme4'<br>ERROR: lazy loading failed for package 'mi'<br>* removing '/mnt/hgfs/bolker/Documents/R/pkgs/lme4/pkgtests/mi.Rcheck/mi'
-</td></tr>
-
-<tr>
-<td class=firstcolumn>36
-</td>
<td class=cellinside>mice
</td>
<td class=cellinside>CRAN
@@ -634,7 +586,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>37
+<td class=firstcolumn>34
</td>
<td class=cellinside>mixAK
</td>
@@ -650,7 +602,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>38
+<td class=firstcolumn>35
</td>
<td class=cellinside>MixMod
</td>
@@ -666,23 +618,23 @@
</td></tr>
<tr>
-<td class=firstcolumn>39
+<td class=firstcolumn>36
</td>
<td class=cellinside>mlmRev
</td>
-<td class=cellinside>CRAN
+<td class=cellinside>local
</td>
-<td class=cellinside>1.0-1
+<td class=cellinside>1.0-3
</td>
<td class=cellinside><lme4-authors at R-forge.wu-wien.ac.at>
</td>
-<td class=cellinside><font style="color:red">error_examples</font>
+<td class=cellinside><font style="color:blue">OK</font>
</td>
-<td class=cellinside>82: Error in if (REML) p else 0L : argument is not interpretable as logical<br>83: Calls: lmer -> mkRespMod<br>84: Execution halted
+<td class=cellinside>
</td></tr>
<tr>
-<td class=firstcolumn>40
+<td class=firstcolumn>37
</td>
<td class=cellinside>multcomp
</td>
@@ -692,13 +644,13 @@
</td>
<td class=cellinside>Torsten Hothorn
</td>
-<td class=cellinside><font style="color:red">error_vignette</font>
+<td class=cellinside><font style="color:blue">OK</font>
</td>
-<td class=cellinside>115: Error: PIRLS step failed<br>116: Execution halted<br>117:
+<td class=cellinside>
</td></tr>
<tr>
-<td class=firstcolumn>41
+<td class=firstcolumn>38
</td>
<td class=cellinside>MuMIn
</td>
@@ -714,7 +666,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>42
+<td class=firstcolumn>39
</td>
<td class=cellinside>NanoStringNorm
</td>
@@ -730,7 +682,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>43
+<td class=firstcolumn>40
</td>
<td class=cellinside>NetworkAnalysis
</td>
@@ -746,7 +698,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>44
+<td class=firstcolumn>41
</td>
<td class=cellinside>odprism
</td>
@@ -758,11 +710,11 @@
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
-<td class=cellinside>167: caught warning: Cholmod warning 'not positive definite' at file:../Cholesky/t_cholmod_rowfac.c, line 432 <br>168: caught warning: Cholmod warning 'not positive definite' at file:../Cholesky/t_cholmod_rowfac.c, line 432 <br>169: Execution halted
+<td class=cellinside>168: caught warning: Cholmod warning 'not positive definite' at file:../Cholesky/t_cholmod_rowfac.c, line 432 <br>169: caught warning: Cholmod warning 'not positive definite' at file:../Cholesky/t_cholmod_rowfac.c, line 432 <br>170: Execution halted
</td></tr>
<tr>
-<td class=firstcolumn>45
+<td class=firstcolumn>42
</td>
<td class=cellinside>ordinal
</td>
@@ -778,7 +730,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>46
+<td class=firstcolumn>43
</td>
<td class=cellinside>pamm
</td>
@@ -794,7 +746,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>47
+<td class=firstcolumn>44
</td>
<td class=cellinside>pbkrtest
</td>
@@ -810,7 +762,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>48
+<td class=firstcolumn>45
</td>
<td class=cellinside>pedigreemm
</td>
@@ -826,7 +778,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>49
+<td class=firstcolumn>46
</td>
<td class=cellinside>phmm
</td>
@@ -838,12 +790,28 @@
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
-<td class=cellinside>132: "ddenseMatrix#dgeMatrix" would also be valid<br>133: Error: trying to get slot "AICtab" from an object (class "summary.mer") that is not an S4 object <br>134: Execution halted
+<td class=cellinside>134: "ddenseMatrix#dgeMatrix" would also be valid<br>135: Error: trying to get slot "AICtab" from an object (class "summary.mer") that is not an S4 object <br>136: Execution halted
</td></tr>
<tr>
-<td class=firstcolumn>50
+<td class=firstcolumn>47
</td>
+<td class=cellinside>polytomous
+</td>
+<td class=cellinside>CRAN
+</td>
+<td class=cellinside>0.1.4
+</td>
+<td class=cellinside>Antti Arppe
+</td>
+<td class=cellinside><font style="color:gray">skipped</font>
+</td>
+<td class=cellinside>
+</td></tr>
+
+<tr>
+<td class=firstcolumn>48
+</td>
<td class=cellinside>prLogistic
</td>
<td class=cellinside>CRAN
@@ -854,11 +822,11 @@
</td>
<td class=cellinside><font style="color:red">error_examples</font>
</td>
-<td class=cellinside>92: PIRLS step failed<br>93: Calls: prLogisticDelta ... stopifnot -> fn -> pwrssUpdate -> .Call -> cpp_exception<br>94: Execution halted
+<td class=cellinside>94: PIRLS step failed<br>95: Calls: prLogisticDelta ... stopifnot -> fn -> pwrssUpdate -> .Call -> cpp_exception<br>96: Execution halted
</td></tr>
<tr>
-<td class=firstcolumn>51
+<td class=firstcolumn>49
</td>
<td class=cellinside>R2admb
</td>
@@ -874,7 +842,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>52
+<td class=firstcolumn>50
</td>
<td class=cellinside>R2STATS
</td>
@@ -886,11 +854,11 @@
</td>
<td class=cellinside><font style="color:red">error_install</font>
</td>
-<td class=cellinside>** preparing package for lazy loading<br>Error in loadNamespace(i, c(lib.loc, .libPaths())) : <br> there is no package called 'RGtk2'<br>Error : package 'gWidgetsRGtk2' could not be loaded<br>ERROR: lazy loading failed for package 'R2STATS'<br>* removing '/mnt/hgfs/bolker/Documents/R/pkgs/lme4/pkgtests/R2STATS.Rcheck/R2STATS'
+<td class=cellinside>** R<br>** inst<br>** preparing package for lazy loading<br>Error : object 'logLik' is not exported by 'namespace:lme4'<br>ERROR: lazy loading failed for package 'R2STATS'<br>* removing '/mnt/hgfs/bolker/Documents/R/pkgs/lme4/pkgtests/R2STATS.Rcheck/R2STATS'
</td></tr>
<tr>
-<td class=firstcolumn>53
+<td class=firstcolumn>51
</td>
<td class=cellinside>RLRsim
</td>
@@ -906,7 +874,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>54
+<td class=firstcolumn>52
</td>
<td class=cellinside>SASmixed
</td>
@@ -922,7 +890,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>55
+<td class=firstcolumn>53
</td>
<td class=cellinside>sdtalt
</td>
@@ -938,7 +906,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>56
+<td class=firstcolumn>54
</td>
<td class=cellinside>TripleR
</td>
@@ -954,7 +922,7 @@
</td></tr>
<tr>
-<td class=firstcolumn>57
+<td class=firstcolumn>55
</td>
<td class=cellinside>Zelig
</td>
@@ -976,7 +944,7 @@
<hr size=1>
<font size=-1>
- Generated on: <i>Tue May 15 13:46:03 2012</i> - <b>R2HTML</b>
+ Generated on: <i>Wed May 16 15:02:30 2012</i> - <b>R2HTML</b>
<hr size=1>
</body>
</html>
\ No newline at end of file
Modified: www/misc/lme4_conversion.html
===================================================================
--- www/misc/lme4_conversion.html 2012-05-16 20:16:32 UTC (rev 1742)
+++ www/misc/lme4_conversion.html 2012-05-16 21:17:42 UTC (rev 1743)
@@ -5,7 +5,9 @@
><h2
>Version numbering</h2
><p
- >As previously announced on the mailing list, we will shortly be releasing a new version of <code
+ >As <a href="https://stat.ethz.ch/pipermail/r-sig-mixed-models/2012q1/014811.html"
+ >previously announced on the lme4 mailing list</a
+ >, we will shortly be releasing a new version of <code
>lme4</code
>, a descendant of the previous development version <code
>lme4Eigen</code
@@ -31,138 +33,156 @@
> from R-forge) are deprecated.</li
></ul
></div
- ><div id="changes-in-behavior"
+ ><div id="release-notes-for-end-users"
><h2
- >Changes in behavior</h2
- ><ul
- ><li
- >Because the internal computational machinery has changed, results from the newest version of <code
- >lme4</code
- > will not be numerically identical to those from previous versions. For reasonably well-defined fits, they will be extremely close (within numerical tolerances of 1e-4 or so), but for unstable or poorly-defined fits the results may change, and very unstable fits may fail when they (apparently) succeeded with previous versions. Similarly, some fits may be slower with the new version, although on average the new version should be faster and more stable. There are more numerical tuning options available than before (see below); non-default settings may restore the speed and/or ability to fit a particular model without an error.</li
+ >Release notes for end-users</h2
+ ><div id="changes-in-behavior"
+ ><h3
+ >Changes in behavior</h3
+ ><ul
><li
- >In the past, <code
- >lmer</code
- > automatically called <code
- >glmer</code
- > when <code
- >family</code
- > was specified. It still does so, but now warns the user that they should preferably use <code
- >glmer</code
- > directly</li
+ >Because the internal computational machinery has changed, results from the newest version of <code
+ >lme4</code
+ > will not be numerically identical to those from previous versions. For reasonably well-defined fits, they will be extremely close (within numerical tolerances of 1e-4 or so), but for unstable or poorly-defined fits the results may change, and very unstable fits may fail when they (apparently) succeeded with previous versions. Similarly, some fits may be slower with the new version, although on average the new version should be faster and more stable. There are more numerical tuning options available than before (see below); non-default settings may restore the speed and/or ability to fit a particular model without an error.</li
+ ><li
+ >In the past, <code
+ >lmer</code
+ > automatically called <code
+ >glmer</code
+ > when <code
+ >family</code
+ > was specified. It still does so, but now warns the user that they should preferably use <code
+ >glmer</code
+ > directly.</li
+ ><li
+ ><code
+ >VarCorr</code
+ > returns its results in the same format as before (as a list of variance-covariance matrices with <code
+ >correlation</code
+ > and <code
+ >stddev</code
+ > attributes, plus a <code
+ >sc</code
+ > attribute giving the residual standard deviation/scale parameter when appropriate), but prints them in a different (nicer) way.</li
+ ><li
+ >by default <code
+ >residuals</code
+ > gives deviance (rather than Pearson) residuals when applied to <code
+ >glmer</code
+ > fits (a side effect of matching <code
+ >glm</code
+ > behaviour more closely).</li
+ ></ul
+ ></div
+ ><div id="other-user-visible-changes"
+ ><h3
+ >Other user-visible changes</h3
+ ><ul
><li
- ><code
- >VarCorr</code
- > returns its results in the same format as before (as a list of variance-covariance matrices with <code
- >correlation</code
- > and <code
- >stddev</code
- > attributes, plus a <code
- >sc</code
- > attribute giving the residual standard deviation/scale parameter when appropriate), but prints them in a different (nicer) way</li
- ></ul
- ></div
- ><div id="other-user-visible-changes"
- ><h2
- >Other user-visible changes</h2
- ><ul
- ><li
- >methods such as <code
- >fixef</code
- > no longer conflict with the <code
- >nlme</code
- > package</li
+ >More use is made of S3 rather than S4 classes and methods: one side effect is that methods such as <code
+ >fixef</code
+ > no longer conflict with the <code
+ >nlme</code
+ > package.</li
+ ><li
+ >The internal optimizer has changed. <code
+ >[gn]lmer</code
+ > now has an <code
+ >optimizer</code
+ > argument; <code
+ >"Nelder_Mead"</code
+ > is the default for <code
+ >[n]lmer</code
+ >, while a combination of <code
+ >"bobyqa"</code
+ > (an alternative derivative-free method) and <code
+ >"Nelder_Mead"</code
+ > is the default for <code
+ >glmer</code
+ >; to use the <code
+ >nlminb</code
+ > optimizer as in the old version of lme4, you can use <code
+ >optimizer="optimx"</code
+ > with <code
+ >control=list(method="nlminb")</code
+ > (you will need the <code
+ >optimx</code
+ > package to be installed and loaded). See the help pages for details.</li
+ ><li
+ >(FIXME: describe residuals)</li
+ ><li
+ >Families in GLMMs are no longer restricted to built-in/hard-coded families; any family described in <code
+ >?family</code
+ >, or following that design, is usable (although there are some hard-coded families, which will be faster)</li
+ ><li
+ ><code
+ >[gn]lmer</code
+ > now produces objects of class <code
+ >merMod</code
+ > rather than class <code
+ >mer</code
+ > as before</li
+ ></ul
+ ></div
+ ><div id="new-features"
+ ><h3
+ >New features</h3
+ ><ul
><li
- ><code
- >[gn]lmer</code
- > now produces objects of class <code
- >merMod</code
- > rather than class <code
- >mer</code
- > as before</li
+ >a general-purpose <code
+ >getME()</code
+ > accessor method has been used to allow extraction of a wide variety of components of a mixed-model fit; this has been backported to <code
+ >lme4.0</code
+ > for compatibility</li
+ ><li
+ ><code
+ >bootMer</code
+ >, a framework for obtaining parameter confidence intervals by parametric bootstrapping</li
+ ><li
+ ><code
+ >plot</code
+ > methods similar to those from the <code
+ >nlme</code
+ > package (although missing <code
+ >augPred</code
+ >)</li
+ ><li
+ >a <code
+ >predict</code
+ > method, allowing a choice of which random effects are included in the prediction</li
+ ><li
+ >likelihood profiling (and profile confidence intervals) for <code
+ >lmer</code
+ > and <code
+ >glmer</code
+ > results</li
+ ><li
+ ><code
+ >nAGQ=0</code
+ >, an option to do fast (but inaccurate) fitting of GLMMs</li
+ ><li
+ >negative binomial models</li
+ ></ul
+ ></div
+ ><div id="still-non-existent-features"
+ ><h3
+ >Still non-existent features</h3
+ ><ul
><li
- ><code
- >[gn]lmer</code
- > now has an <code
- >optimizer</code
- > argument. <code
- >"Nelder_Mead"</code
- > is the default for <code
- >[n]lmer</code
- >, while a combination of <code
- >"bobyqa"</code
- > (an alternative derivative-free method) and <code
- >"Nelder_Mead"</code
- > is the default for <code
- >glmer</code
- >; to use the <code
- >nlminb</code
- > optimizer as in the old version of lme4, you can use <code
- >optimizer="optimx"</code
- > with <code
- >control=list(method="nlminb")</code
- > (you will need the <code
- >optimx</code
- > package to be installed and loaded). See the help pages for details.</li
- ><li
- >(FIXME: describe residuals)</li
- ></ul
+ >Automatic MCMC sampling based on the fit turns out to be very difficult to implement in a way that is really broadly reliable and robust; <code
+ >mcmcsamp</code
+ > will not be implemented in the near future</li
+ ><li
+ >"R-side" structures (within-block correlation and heteroscedasticity) are not on the current timetable</li
+ ></ul
+ ></div
></div
- ><div id="new-features"
+ ><div id="notes-for-package-writers"
><h2
- >New features</h2
- ><ul
- ><li
- >a general-purpose <code
- >getME()</code
- > accessor method has been used to allow extraction of a wide variety of components of a mixed-model fit; this has been backported to <code
- >lme4.0</code
- > for compatibility</li
- ><li
- ><code
- >bootMer</code
- >, a framework for obtaining parameter confidence intervals by parametric bootstrapping</li
- ><li
- ><code
- >plot</code
- > methods similar to those from the <code
- >nlme</code
- > package (although without <code
- >augPred</code
- >)</li
- ><li
- >a <code
- >predict</code
- > method, allowing a choice of which random effects are included in the prediction</li
- ><li
- >profiling (and profile confidence intervals) for <code
- >lmer</code
- > and <code
- >glmer</code
- > results</li
- ><li
- ><code
- >nAGQ=0</code
- >, an option to do fast (but inaccurate) fitting of GLMMs</li
- ></ul
- ></div
- ><div id="still-non-existent-features"
- ><h2
- >Still non-existent features</h2
- ><ul
- ><li
- >Automatic MCMC sampling based on the fit turns out to be very difficult to implement in a way that is really broadly reliable and robust; <code
- >mcmcsamp</code
- > will not be implemented in the near future</li
- ><li
- >"R-side" structures (within-block correlation and heteroscedasticity) are not on the current timetable</li
- ></ul
- ></div
- ><div id="for-package-writers:"
- ><h2
- >For package writers:</h2
+ >Notes for package writers</h2
><p
><a href="./lme4_compat_report.html"
- >Current package compatibilty test results</a
+ >Current package compatibility test results</a
></p
><ul
><li
@@ -170,15 +190,19 @@
>lme4</code
>-old and <code
>lme4.0</code
- > produces <code
+ > produce objects of class <code
>mer</code
- > objects, <code
+ >, <code
>lme4</code
>-new produces <code
>merMod</code
- > objects</li
+ > objects, so any methods written for class <code
+ >mer</code
+ > will at least have to be copied to work with class <code
+ >merMod</code
+ ></li
><li
- >you can distinguish <code
+ >You can distinguish <code
>lme4</code
>-old from <code
>lme4</code
@@ -190,32 +214,32 @@
><code
>http://lme4.r-forge.r-project.org/repos</code
></a
- > is 0.999902344-0)</li
+ > is 0.999902344-0).</li
><li
- >so you can test e.g. if <code
+ >So you can test e.g. if <code
>packageVersion("lme4")<="0.999375-43"</code
- > (yes, you do want the quotation marks; package versions are weird objects in R)</li
+ > (yes, you do want the quotation marks; package versions are weird objects in R).</li
><li
><code
>getME(.,.)</code
> should often get the components you want</li
><li
- ><code
+ >As mentioned above, <code
>lme4</code
- > now uses S3 rather than S4 methods in many places. That makes some things easier ...</li
+ > now uses S3 rather than S4 methods in many places. That makes some things easier, but it also means (for example) that classes and methods do not get imported in the same way.</li
></ul
- ><div id="things-that-wont-work:"
+ ><div id="things-that-wont-work"
><h3
- >Things that won't work:</h3
+ >Things that won't work</h3
><ul
><li
- >direct extraction of slots via <code
+ >Direct extraction of slots via <code
>@</code
> (use <code
>getME()</code
- > instead)</li
+ > instead).</li
><li
- >methods that depend on <code
+ >Methods that depend on <code
>lme4</code
> producing objects of class <code
>mer</code
@@ -229,9 +253,9 @@
>isGLMM()</code
>, <code
>isNLMM()</code
- > that should help you distinguish different types of model if you need to)</li
+ > that should help you distinguish different types of model if you need to).</li
><li
- >the <code
+ >The <code
>method</code
> argument is no longer used in <code
>glmer</code
@@ -239,19 +263,17 @@
>nAGQ=1</code
> vs <code
>nAGQ>1</code
- > specifies whether to use Laplace or</li
+ > specifies whether to use Laplace or AGQ).</li
><li
><code
>expandSlash</code
- > no longer exists (<em
- >although it does exist within the <code
- >findbars</code
- > function: could be reconstituted??</em
- >)</li
+ > no longer exists (although it does exist within the <code
+ >findbars</code
+ > function: could be reconstituted??).</li
><li
- >because S4 methods are used less, and (S4) <em
+ >Because S4 methods are used less, and (S4) <em
>reference</em
- > classes are used, considerably fewer methods are exported, but they are generally available as reference class method "slots"</li
+ > classes are used, considerably fewer methods are exported, but they are generally available as reference class method "slots".</li
></ul
></div
></div
Modified: www/misc/lme4_conversion.md
===================================================================
--- www/misc/lme4_conversion.md 2012-05-16 20:16:32 UTC (rev 1742)
+++ www/misc/lme4_conversion.md 2012-05-16 21:17:42 UTC (rev 1743)
@@ -2,51 +2,58 @@
## Version numbering
-As previously announced on the mailing list, we will shortly be releasing a new version of `lme4`, a descendant of the previous development version `lme4Eigen`. For users who do not access any internal structures, there will be few backward-incompatible changes.
+As [previously announced on the lme4 mailing list][announce], we will shortly be releasing a new version of `lme4`, a descendant of the previous development version `lme4Eigen`. For users who do not access any internal structures, there will be few backward-incompatible changes.
+[announce]: https://stat.ethz.ch/pipermail/r-sig-mixed-models/2012q1/014811.html
+
* the version of `lme4` currently on R-forge (currently version 0.999902344-0, to be released as version 1.0 (!)) should be used for new projects
* The current CRAN version (0.999375-42) will be replaced by a nearly identical version called `lme4.0` (currently version 0.9999-2; the only backward-incompatible change in this version is a fix to the AGQ code contributed by Wayne Zhang). `lme4.0` is a maintenance version and will only be changed to fix documented bugs.
* all other versions (`lme4a`, `lme4b`, `lme4Eigen` from R-forge) are deprecated.
-## Changes in behavior
+## Release notes for end-users
+
+### Changes in behavior
* Because the internal computational machinery has changed, results from the newest version of `lme4` will not be numerically identical to those from previous versions. For reasonably well-defined fits, they will be extremely close (within numerical tolerances of 1e-4 or so), but for unstable or poorly-defined fits the results may change, and very unstable fits may fail when they (apparently) succeeded with previous versions. Similarly, some fits may be slower with the new version, although on average the new version should be faster and more stable. There are more numerical tuning options available than before (see below); non-default settings may restore the speed and/or ability to fit a particular model without an error.
-* In the past, `lmer` automatically called `glmer` when `family` was specified. It still does so, but now warns the user that they should preferably use `glmer` directly
-* `VarCorr` returns its results in the same format as before (as a list of variance-covariance matrices with `correlation` and `stddev` attributes, plus a `sc` attribute giving the residual standard deviation/scale parameter when appropriate), but prints them in a different (nicer) way
+* In the past, `lmer` automatically called `glmer` when `family` was specified. It still does so, but now warns the user that they should preferably use `glmer` directly.
+* `VarCorr` returns its results in the same format as before (as a list of variance-covariance matrices with `correlation` and `stddev` attributes, plus a `sc` attribute giving the residual standard deviation/scale parameter when appropriate), but prints them in a different (nicer) way.
+* by default `residuals` gives deviance (rather than Pearson) residuals when applied to `glmer` fits (a side effect of matching `glm` behaviour more closely).
-## Other user-visible changes
-* methods such as `fixef` no longer conflict with the `nlme` package
+### Other user-visible changes
+* More use is made of S3 rather than S4 classes and methods: one side effect is that methods such as `fixef` no longer conflict with the `nlme` package.
+* The internal optimizer has changed. `[gn]lmer` now has an `optimizer` argument; `"Nelder_Mead"` is the default for `[n]lmer`, while a combination of `"bobyqa"` (an alternative derivative-free method) and `"Nelder_Mead"` is the default for `glmer`; to use the `nlminb` optimizer as in the old version of lme4, you can use `optimizer="optimx"` with `control=list(method="nlminb")` (you will need the `optimx` package to be installed and loaded). See the help pages for details.
+* (FIXME: describe residuals)
+* Families in GLMMs are no longer restricted to built-in/hard-coded families; any family described in `?family`, or following that design, is usable (although there are some hard-coded families, which will be faster)
* `[gn]lmer` now produces objects of class `merMod` rather than class `mer` as before
-* `[gn]lmer` now has an `optimizer` argument. `"Nelder_Mead"` is the default for `[n]lmer`, while a combination of `"bobyqa"` (an alternative derivative-free method) and `"Nelder_Mead"` is the default for `glmer`; to use the `nlminb` optimizer as in the old version of lme4, you can use `optimizer="optimx"` with `control=list(method="nlminb")` (you will need the `optimx` package to be installed and loaded). See the help pages for details.
-* (FIXME: describe residuals)
-## New features
+### New features
* a general-purpose `getME()` accessor method has been used to allow extraction of a wide variety of components of a mixed-model fit; this has been backported to `lme4.0` for compatibility
* `bootMer`, a framework for obtaining parameter confidence intervals by parametric bootstrapping
-* `plot` methods similar to those from the `nlme` package (although without `augPred`)
+* `plot` methods similar to those from the `nlme` package (although missing `augPred`)
* a `predict` method, allowing a choice of which random effects are included in the prediction
-* profiling (and profile confidence intervals) for `lmer` and `glmer` results
+* likelihood profiling (and profile confidence intervals) for `lmer` and `glmer` results
* `nAGQ=0`, an option to do fast (but inaccurate) fitting of GLMMs
+* negative binomial models
-## Still non-existent features
+### Still non-existent features
* Automatic MCMC sampling based on the fit turns out to be very difficult to implement in a way that is really broadly reliable and robust; `mcmcsamp` will not be implemented in the near future
* "R-side" structures (within-block correlation and heteroscedasticity) are not on the current timetable
-## For package writers:
+## Notes for package writers
-[Current package compatibilty test results][pkgtest]
+[Current package compatibility test results][pkgtest]
[pkgtest]: ./lme4_compat_report.html
-* `lme4`-old and `lme4.0` produces `mer` objects, `lme4`-new produces `merMod` objects
-* you can distinguish `lme4`-old from `lme4`-new via package version; the last old-style version of `lme4` on CRAN is 0.999375-42, so anything after that is `lme4`-new (the current version on <http://lme4.r-forge.r-project.org/repos> is 0.999902344-0)
-* so you can test e.g. if `packageVersion("lme4")<="0.999375-43"` (yes, you do want the quotation marks; package versions are weird objects in R)
+* `lme4`-old and `lme4.0` produce objects of class `mer`, `lme4`-new produces `merMod` objects, so any methods written for class `mer` will at least have to be copied to work with class `merMod`
+* You can distinguish `lme4`-old from `lme4`-new via package version; the last old-style version of `lme4` on CRAN is 0.999375-42, so anything after that is `lme4`-new (the current version on <http://lme4.r-forge.r-project.org/repos> is 0.999902344-0).
+* So you can test e.g. if `packageVersion("lme4")<="0.999375-43"` (yes, you do want the quotation marks; package versions are weird objects in R).
* `getME(.,.)` should often get the components you want
-* `lme4` now uses S3 rather than S4 methods in many places. That makes some things easier ...
+* As mentioned above, `lme4` now uses S3 rather than S4 methods in many places. That makes some things easier, but it also means (for example) that classes and methods do not get imported in the same way.
-### Things that won't work:
-* direct extraction of slots via `@` (use `getME()` instead)
-* methods that depend on `lme4` producing objects of class `mer` (write new methods for class `merMod`; `lme4` now has `isLMM()`, `isGLMM()`, `isNLMM()` that should help you distinguish different types of model if you need to)
-* the `method` argument is no longer used in `glmer` (`nAGQ=1` vs `nAGQ>1` specifies whether to use Laplace or
[TRUNCATED]
To get the complete diff run:
svnlook diff /svnroot/lme4 -r 1743
More information about the Lme4-commits
mailing list