Permalink
Switch branches/tags
Nothing to show
Find file Copy path
2e1f4dc Sep 26, 2018
2 contributors

Users who have contributed to this file

@bbolker @sjPlot
1041 lines (794 sloc) 85.9 KB
---
author: Ben Bolker and others
title: GLMM FAQ
bibliography: glmm.bib
date: "`r format(Sys.time(), '%H:%M %d %B %Y')`"
output:
html_document:
toc: true
---
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-3131047-9"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-3131047-9');
</script>
```{r setup,message=FALSE,echo=FALSE}
library(knitr)
library(pander)
```
# Introduction
This is an informal FAQ list for the `r-sig-mixed-models` mailing list.
The most commonly used functions for mixed modeling in R are
- *linear mixed models*: `aov()`, `nlme::lme`[^1], `lme4::lmer`; `brms::brm`
- *generalized linear mixed models* (GLMMs)
- frequentist: `MASS::glmmPQL`, `lme4::glmer`; `glmmTMB`
- Bayesian: `MCMCglmm::MCMCglmm`; `brms::brm`
- *nonlinear mixed models*: `nlme::nlme`, `lme4::nlmer`; `brms::brm`
- *GNLMMs*: `brms::brm`
[^1]: in R, `foo::bar` (or `foo::bar()`) denotes "function `bar` in package `foo`").
Another quick-and-dirty way to search for mixed-model related packages on CRAN:
```{r avail}
grep("l.?m[me][^t]",rownames(available.packages()),value=TRUE)
```
## Other sources of help
- the mailing list is `r-sig-mixed-models@r-project.org`
- sign up [here](https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models)
- archives [here](https://stat.ethz.ch/pipermail/r-sig-mixed-models/)
- or Google search with the tag `site:https://stat.ethz.ch/pipermail/r-sig-mixed-models/`
- The source code of this document is available [on GitHub](https://github.com/bbolker/mixedmodels-misc/blob/master/glmmFAQ.rmd); the rendered (HTML) version lives on [GitHub pages](http://bbolker.github.io/mixedmodels-misc/glmmFAQ.html).
- Searching on StackOverflow with the [[r] [mixed-models] tags](http://stackoverflow.com/questions/tagged/r%20mixed-models?mode=all), or on CrossValidated with the [[mixed-model] tag](http://stats.stackexchange.com/questions/tagged/mixed-model) may be helpful (these sites also have an `[lme4]` tag).
**DISCLAIMERS:**
- (G)LMMs are hard - harder than you may think based on what you may have learned in your second statistics class, which probably focused on picking the appropriate sums of squares terms and degrees of freedom for the numerator and denominator of an $F$ test. 'Modern' mixed model approaches, although more powerful (they can handle more complex designs, lack of balance, crossed random factors, some kinds of non-Normally distributed responses, etc.), also require a new set of conceptual tools. In order to use these tools you should have at least a general acquaintance with classical mixed-model experimental designs but you should also, probably, read something about modern mixed model approaches. @littell_sas_2006 and @pinheiro_mixed-effects_2000 are two places to start, although Pinheiro and Bates is probably more useful if you want to use R. Other useful references include @gelman_data_2006 (focused on Bayesian methods) and @zuur_mixed_2009. If you are going to use generalized linear mixed models, you should understand generalized linear models (@dobson_introduction_2008, @faraway_extending_2006, and @McCullaghNelder1989 are standard references; the last is the canonical reference, but also the most challenging).
- All of the issues that arise with regular linear or generalized-linear modeling (e.g.: inadequacy of p-values alone for thorough statistical analysis; need to understand how models are parameterized; need to understand the principle of marginality and how interactions can be treated; dangers of overfitting, which are not mitigated by stepwise procedures; the non-existence of free lunches) also apply, and can apply more severely, to mixed models.
* When SAS (or Stata, or Genstat/AS-REML or ...) and R differ in their answers, R may not be wrong. Both SAS and R may be `right' but proceeding in a different way/answering different questions/using a different philosophical approach (or both may be wrong ...)
- The advice in this FAQ comes with **absolutely no warranty of any sort**.
# Model definition
## Model specification
The following formula extensions for specifying random-effects structures in R are used by
- `lme4`
- `nlme` (nested effects only, although crossed effects can be specified with more work)
- `glmmADMB` and `glmmTMB`
`MCMCglmm` uses a different specification, inherited from AS-REML.
(Modified from Robin Jeffries, UCLA:)
```{r glmm_syntax,echo=FALSE}
spectab <- read.table(sep="&",header=TRUE,text="
formula & meaning
`(1|group)` & random group intercept
`(x|group)` = `(1+x|group)` & random slope of x within group with correlated intercept
`(0+x|group)` = `(-1+x|group)` & random slope of x within group: no variation in intercept
`(1|group) + (0+x|group)` & uncorrelated random intercept and random slope within group
`(1|site/block)` = `(1|site)+(1|site:block)` & intercept varying among sites and among blocks within sites (nested random effects)
`site+(1|site:block)` & *fixed* effect of sites plus random variation in intercept among blocks within sites
`(x|site/block)` = `(x|site)+(x|site:block)` = `(1 + x|site)+(1+x|site:block)` & slope and intercept varying among sites and among blocks within sites
`(x1|site)+(x2|block)` & two different effects, varying at different levels
`x*site+(x|site:block)` & fixed effect variation of slope and intercept varying among sites and random variation of slope and intercept among blocks within sites
`(1|group1)+(1|group2)` & intercept varying among crossed random effects (e.g. site, year)")
set.alignment(default=c("centre","left"))
pander(spectab)
```
Two more possibly useful links:
- [Rense Nieuwenhuis's blogpost/lesson on lme4 model specification](http://www.rensenieuwenhuis.nl/r-sessions-16-multilevel-model-specification-lme4/)
- CrossValidated's [lmer cheat sheet](https://stats.stackexchange.com/questions/13166/rs-lmer-cheat-sheet)
## Should I treat factor xxx as fixed or random?
This is in general a far more difficult question than it seems on the surface. There are many competing philosophies and definitions. For example, from @gelman_analysis_2005:
> Before discussing the technical issues, we briefly review what is meant by fixed and random effects. It turns out that different—in fact, incompatible—definitions are used in different contexts. [See also Kreft and de Leeuw (1998), Section 1.3.3, for a discussion of the multiplicity of definitions of fixed and random effects and coefficients, and Robinson (1998) for a historical overview.] Here we outline five definitions that we have seen: 1. Fixed effects are constant across individuals, and random effects vary. For example, in a growth study, a model with random intercepts αi and fixed slope β corresponds to parallel lines for different individuals i, or the model yit = αi + βt. Kreft and de Leeuw [(1998), page 12] thus distinguish between fixed and random coefficients. 2. Effects are fixed if they are interesting in themselves or random if there is interest in the underlying population. Searle, Casella and McCulloch [(1992), Section 1.4] explore this distinction in depth. 3. “When a sample exhausts the population, the corresponding variable is fixed; when the sample is a small (i.e., negligible) part of the population the corresponding variable is random” [Green and Tukey (1960)]. 4. “If an effect is assumed to be a realized value of a random variable, it is called a random effect” [LaMotte (1983)]. 5. Fixed effects are estimated using least squares (or, more generally, maximum likelihood) and random effects are estimated with shrinkage [“linear unbiased prediction” in the terminology of Robinson (1991)]. This definition is standard in the multilevel modeling literature [see, e.g., Snijders and Bosker (1999), Section 4.2] and in econometrics.
Another useful comment (via Kevin Wright) reinforcing the idea that "random vs. fixed" is not a simple, cut-and-dried decision: from @schabenberger_contemporary_2001, p. 627:
> Before proceeding further with random field linear models we need to remind the reader of the adage that one modeler's random effect is another modeler's fixed effect.
@clark2015should address this question from a mostly econometric perspective, focusing mostly on practical variance/bias/RMSE criteria.
One point of particular relevance to 'modern' mixed model estimation (rather than 'classical' method-of-moments estimation) is that, for practical purposes, there must be a reasonable number of random-effects levels (e.g. blocks) -- more than 5 or 6 at a minimum. This is not surprising if you consider that random effects estimation is trying to estimate an among-block variance. For example, from @Crawley2002 p. 670:
> Are there enough levels of the factor in the data on which to base an estimate of the variance of the population of effects? No, means [you should probably treat the variable as] fixed effects.
Some researchers (who treat fixed vs random as a philosophical rather than a pragmatic decision) object to this approach.
Treating factors with small numbers of levels as random will in the best case lead to very small and/or imprecise estimates of random effects; in the worst case it will lead to various numerical difficulties such as lack of convergence, zero variance estimates, etc.. (A small [simulation exercise](https:/rpubs.com/bbolker/4187) shows that at least the estimates of the standard deviation are downwardly biased in this case; it's not clear whether/how this bias would affect the point estimates of fixed effects or their estimated confidence intervals.) In the classical method-of-moments approach these problems may not arise (because the sums of squares are always well defined as long as there are at least two units), but the underlying problems of lack of power are there nevertheless.
Also see [this thread](https://stat.ethz.ch/pipermail/r-sig-mixed-models/2010q2/003709.html) on the r-sig-mixed-models mailing list.
## Nested or crossed?
<a id="nested_or_crossed"></a>
- Relatively few mixed effect modeling packages can handle crossed random effects, i.e. those where one level of a random effect can appear in conjunction with more than one level of another effect. (This definition is confusing, and I would happily accept a better one.) A classic example is crossed temporal and spatial effects. If there is random variation among temporal blocks (e.g. years) ''and'' random variation among spatial blocks (e.g. sites), ''and'' if there is a consistent year effect across sites and ''vice versa'', then the random effects should be treated as crossed.
- `lme4` does handled crossed effects, efficiently; if you need to deal with crossed REs in conjunction with some of the features that `nlme` offers (e.g. heteroscedasticity of residuals via `weights`/`varStruct`, correlation of residuals via `correlation`/`corStruct`, see p. 163ff of @pinheiro_mixed-effects_2000 (section 4.2.2: [Google books link](http://tinyurl.com/crossedRE))
- I rarely find it useful to think of fixed effects as "nested" (although others disagree); if for example treatments A and B are only measured in block 1, and treatments C and D are only measured in block 2, one still assumes (because they are fixed effects) that each treatment would have the same effect if applied in the other block. (One might like to estimate treatment-by-block interactions, but in this case the experimental design doesn't allow it; one would have to have multiple treatments measured within each block, although not necessarily all treatments in every block.) One would code this analysis as `response~treatment+(1|block)` in `lme4`. Also, in the case of fixed effects, crossed and nested specifications change the parameterization of the model, but not anything else (e.g. the number of parameters estimated, log-likelihood, model predictions are all identical). That is, in R's `model.matrix` function (which implements a version of Wilkinson-Rogers notation) `a*b` and `a/b` (which expand to `1+a+b+a:b` and `1+a+a:b` respectively) give model matrices with the same number of columns.
- Whether you explicitly specify a random effect as nested or not depends (in part) on the way the levels of the random effects are coded. If the 'lower-level' random effect is coded with unique levels, then the two syntaxes `(1|a/b)` (or `(1|a+a:b)`) and `(1|a)+(1|b)` are equivalent. If the lower-level random effect has the same labels within each larger group (e.g. blocks 1, 2, 3, 4 within sites A, B, and C) then the explicit nesting `(1|a/b)` is required. It seems to be considered best practice to code the nested level uniquely (e.g. A1, A2, ..., B1, B2, ...) so that confusion between nested and crossed effects is less likely.
# Model extensions (overdispersion, zero-inflation, ...)
## Overdispersion
<a id="overdispersion"></a>
### Testing for overdispersion/computing overdispersion factor
- with the usual caveats, plus a few extras -- counting degrees of freedom, etc. -- the usual procedure of calculating the sum of squared Pearson residuals and comparing it to the residual degrees of freedom should give at least a crude idea of overdispersion. The following attempt counts each variance or covariance parameter as one model degree of freedom and presents the sum of squared Pearson residuals, the ratio of (SSQ residuals/rdf), the residual df, and the $p$-value based on the (approximately!!) appropriate $\chi^2$ distribution. **Do PLEASE note the usual, and extra, caveats noted here: this is an APPROXIMATE estimate of an overdispersion parameter**. Even in the GLM case, the expected deviance per point equaling 1 is only true as the distribution of individual deviates approaches normality, i.e. the usual $\lambda>5$ rules of thumb for Poisson values and $\textrm{min}(Np, N(1-p)) > 5$ for binomial values (e.g. see @venables_modern_2002, [p. 209](http://books.google.com/books?id=974c4vKurNkC&pg=PA209)). (And that's without the extra complexities due to GLMM, i.e. the "effective" residual df should be large enough to make the sums of squares converge on a $\chi^2$ distribution ...)
- Remember that (1) overdispersion is irrelevant for models that estimate a scale parameter (i.e. almost anything but Poisson or binomial: Gaussian, Gamma, negative binomial ...) and (2) overdispersion is not estimable (and hence practically irrelevant) for Bernoulli models (= binary data = binomial with $N=1$).
The following function should work for a variety of model types
(at least `glmmADMB`, `glmmTMB`, `lme4`, ...).
```{r overdisp_fun}
overdisp_fun <- function(model) {
rdf <- df.residual(model)
rp <- residuals(model,type="pearson")
Pearson.chisq <- sum(rp^2)
prat <- Pearson.chisq/rdf
pval <- pchisq(Pearson.chisq, df=rdf, lower.tail=FALSE)
c(chisq=Pearson.chisq,ratio=prat,rdf=rdf,p=pval)
}
```
Example:
```{r lme4,message=FALSE}
library(lme4)
```
```{r overdisp_ex,cache=TRUE,message=FALSE}
set.seed(101)
d <- data.frame(y=rpois(1000,lambda=3),x=runif(1000),
f=factor(sample(1:10,size=1000,replace=TRUE)))
m1 <- glmer(y~x+(1|f),data=d,family=poisson)
overdisp_fun(m1)
library(glmmADMB) ## 0.7.7
m2 <- glmmadmb(y~x+(1|f),data=d,family="poisson")
overdisp_fun(m2)
```
The `gof` function in the `aods3` provides similar functionality
(it reports both deviance- and $\chi^2$-based estimates of
overdispersion and tests).
### Fitting models with overdispersion?
- quasilikelihood estimation: [MASS::glmmPQL](http://finzi.psych.upenn.edu/R/library/MASS/html/glmmPQL.html). Quasi- was deemed unreliable in `lme4`, and is no longer available. (Part of the problem was questionable numerical results in some cases; the other problem was that DB felt that he did not have a sufficiently good understanding of the theoretical framework that would explain what the algorithm was actually estimating in this case.) [geepack::geelgm](http://finzi.psych.upenn.edu/R/library/geepack/html/geeglm.html) may be workable (haven't tried it)
If you really want quasi-likelihood analysis for `glmer` fits, you can do it yourself by adjusting the
coefficient table - i.e., by multiplying the standard error by the square root
of the dispersion factor [^2] and recomputing the $Z$- and $p$-values
accordingly, as follows:
[^2]: the dispersion factor is estimated on a
variance, so we need to take the square root to apply it to the
standard error
```{r quasi}
## extract summary table; you may also be able to do this via
## broom::tidy or broom.mixed::tidy
cc <- coef(summary(m1))
phi <- overdisp_fun(m1)["ratio"]
cc <- within(as.data.frame(cc),
{ `Std. Error` <- `Std. Error`*sqrt(phi)
`z value` <- Estimate/`Std. Error`
`Pr(>|z|)` <- 2*pnorm(abs(`z value`), lower.tail=FALSE)
})
printCoefmat(cc,digits=3)
```
(In this case it doesn't make much difference, since the data we simulated in the first place were Poisson.) Keep in mind that once you switch to quasi-likelihood you will either have to eschew inferential methods such as the likelihood ratio test, profile confidence intervals, AIC, etc., or make more heroic assumptions to compute "quasi-" analogs of all of the above (such as QAIC).
- observation-level random effects (OLRE: this approach should work in most packages). If you want to a citation for this approach, try @elston_analysis_2001, who cite @lawson_disease_1999; apparently there is also an example in section 10.5 of @maindonald_data_2010, and (according to an R-sig-mixed-models post) this is also discussed by @rabehesketh_multilevel_2008. Also see @browne_variance_2005 for an example in the binomial context (i.e. logit-normal-binomial rather than lognormal-Poisson). Agresti's excellent (2002) book @agresti_categorical_2002 also discusses this (section 13.5), referring back to @breslow_extrapoisson_1984 and @hinde_compound_1982. [**Notes**: (a) I haven't checked all these references myself, (b) I can't find the reference any more, but I have seen it stated that observation-level random effect estimation is probably dodgy for PQL approaches as used in Elston et al 2001]
- alternative distributions
- Poisson-lognormal model for counts or binomial-logit-Normal model for proportions (see above, "observation-level random effects")
- negative binomial for counts or beta-binomial for proportions
- `lme4::glmer.nb()` should fit a negative binomial, although it is somewhat slow and fragile compared to some of the other methods suggested here. `lme4` cannot fit beta-binomial models (these cannot be formulated as a part of the exponential family of distributions)
- [glmmADMB](http://github.com/bbolker/glmmADMB/) and [glmmTMB](https://github.com/glmmtmb/glmmTMB/) (both on GitHub) will fit two parameterizations of the negative binomial: `family="nbinom"` (or `family="nbinom2"` in `glmmTMB`) gives the classic parameterization with $\sigma^2=\mu(1+mu/k)$ ("NB2" in Hardin and Hilbe's terminology) while `family="nbinom1"` gives a parameterization with $\sigma^2=\phi \mu$, $\phi>1$ ("NB1" to Hardin and Hilbe). The latter might also be called a "quasi-Poisson" parameterization because it matches the mean-variance relationship assumed by quasi-Poisson models, i.e. the variance is strictly proportional to the mean (although the proportionality constant must be >1, a limitation that does not apply to quasi-likelihood approaches).
- `glmmTMB` allows beta-binomial models ([@harrison_comparison_2015] suggests comparing beta-binomial with OLRE models to assess reliability)
- the `brms` package has a `negbinomial` family (no beta-binomial, but it does have a wide range of other families)
- other packages/approaches (less widely used, or requiring a bit more effort)
- `gamlss.mx:gamlssNP`
- WinBUGS/JAGS (via R2WinBUGS/Rjags)
- AD Model Builder (possibly via `R2admb` package) or `TMB`
- `gnlmm` in the `repeated` package ([off-CRAN](http://www.commanster.eu/rcode.html))
* [ASREML](http://www.asreml.com/software/genstat/htmlhelp/server/GLMM.htm)
Negative binomial models in `glmmTMB` and lognormal-Poisson models in `glmer` (or `MCMCglmm`) are probably the best quick alternatives for overdispersed count data. If you need to explore alternatives (different variance-mean relationships, different distributions), then `ADMB`, `TMB`, `WinBUGS`, `Stan`, `NIMBLE` are the most flexible alternatives.
### Underdispersion
Underdispersion (much *less* variability than expected) is a less common problem than overdispersion.
- mild overdispersion is sometimes ignored, since it tends in general to lead to conservative rather than anti-conservative results
- quasi-likelihood (and the quasi-hack listed above) can handle under- as well as underdispersion
- some other solutions exist, but are less widely implemented
- for distributions with a small range (e.g. litter sizes of large mammals), one can treat responses as ordinal (e.g. using the `ordinal` package, or `MCMCglmm` or `brms` for Bayesian solutions)
- the COM-Poisson distribution and generalized Poisson distributions, implemented in `glmmTMB`, can handle underdispersion (J. Hilbe recommends the latter in [this CrossValidated answer](https://stats.stackexchange.com/questions/67385/what-is-the-appropriate-model-for-underdispersed-count-data). (`VGAM` has a generalized Poisson distribution, but doesn't handle random effects.)
## Gamma GLMMs
While one (well, OK I) would naively think that GLMMs with Gamma distributions would be just as easy (or hard) as any other sort of GLMMs, it seems that they are in fact harder to implement. Basic simulated examples of Gamma GLMMs can fail in lme4 despite analogous problems with Poisson, binomial, etc. distributions. Solutions:
- the default inverse link seems particularly problematic; try other links (especially `family=Gamma(link="log")`) if that is possible/makes sense
- consider whether a lognormal model (i.e. a regular LMM on logged data) would work/makes sense (@lo_transform_2015 argue that the Gamma family with an *identity* link is superior to lognormal models for reaction-time data)
Gamma models can be fitted by a wide variety of platforms (`lme4::glmer`, `MASS::glmmPQL`, `glmmADMB`, `glmmTMB`, `MixedModels.jl`, `MCMCglmm`, `brms` ... not sure about others.
## Zero-inflation
See e.g. @martin_zero_2005 or @warton_many_2005 ("many zeros does not mean zero inflation") or @zuur_zero-truncated_2009 for general information on zero-inflation.
### Count data
- `MCMCglmm` handles zero-truncated, zero-inflated, and zero-altered models, although specifying the models is a little bit tricky: see Sections 5.3 to 5.5 of the [CourseNotes vignette](https://cran.r-project.org/web/packages/MCMCglmm/vignettes/CourseNotes.pdf)
- `glmmADMB` handles
- zero-inflated models (with a single zero-inflation parameter -- i.e., the level of zero-inflation is assumed constant across the whole data set)
- truncated Poisson and negative binomial distributions (which allows two-stage fitting of hurdle models)
- `glmmTMB` handles a variety of Z-I and Z-T models (allows covariates, and random effects, in the zero-alteration model)
- `brms` does too
- Gavin Simpson has a [detailed writeup](http://www.fromthebottomoftheheap.net/2017/05/04/compare-mgcv-with-glmmTMB/) showing that `mgcv::gam()` can do simple mixed models (Poisson, not NB) with zero-inflation, and comparing `mgcv` with `glmmTMB` results
- `gamlssNP` in the `gamlss.mx` package should handle zero-inflation, and the `gamlss.tr` package should handle truncated (i.e. hurdle) models -- but I haven't tried them
- roll-your-own: ADMB/R2admb, WinBUGS/R2WinBUGS, TMB, Stan, ...
### Continuous data
Continuous data are a special case where the mixture model may make less sense.
- The simplest solution is to fit a Bernoulli model to the zero/non-zero data, then a continuous model (lognormal? Gamma?) for the non-zero values; this is effectively a *hurdle model*. Many of the distributions in the exponential family (Gamma; log-Normal; Beta) don't allow zero values anyway, so any observation of zero must be a "structural" zero rather than a "sample" zero.
- you might have *censored* data where all values below a certain limit (e.g. a detection limit) are recorded as zero; in this case you might be able to use `survreg()` and `frailty()` in the `survival` package for random-intercept models (as [suggested on r-help by Thomas Lumley in 2003](https://stat.ethz.ch/pipermail/r-help/2003-June/034905.html) or [on StackOverflow by user 42- in 2014](https://stackoverflow.com/questions/25317021/interaction-terms-and-random-effects-in-tobit-regression-model-in-r).
- The `cplm` package handles 'Tweedie compound Poisson linear models', which in a particular range of parameters allows for skewed continuous responses with a spike at zero
### Tests for zero-inflation
- you can use a likelihood ratio test between the regular and zero-inflated version of the model, but be aware of boundary issues (search "boundary" elsewhere on this page ...) -- the null value (no zero inflation) is on the boundary of the feasible space
- you can use AIC or variations, with the same caveats
- you can use Vuong's test, which is often recommended for testing zero-inflation in GLMs, because under some circumstances the various model flavors under consideration (hurdle vs zero-inflated vs "vanilla") are not nested. Vuong's test is implemented (and referenced) in the `pscl` package, and should be feasible to implement for GLMMs, but I don't know of an implementation. Someone should let me (BMB) know if they find one.
- two untested but reasonable approaches:
- use a `simulate()` method if it exists to construct a simulated distribution of the proportion of zeros expected overall from your model, and compare it to the observed proportion of zeros in the data set
- compute the probability of a zero for each observation. On the basis of (conditionally) independent Bernoulli trials, compute the expected number of zeros and the confidence intervals -- compare it with the observed number.
## Spatial and temporal correlation models, heteroscedasticity ("R-side" models)
In `nlme` these so-called **R-side** (R for "residual") structures are accessible via the `weights`/`VarStruct` (heteroscedasticity) and `correlation`/`corStruct` (spatial or temporal correlation) arguments and data structures. This extension is a bit harder than it might seem. In LMMs it is a natural extension to allow the residual error terms to be components of a single multivariate normal draw; if that MVN distribution is uncorrelated and homoscedastic (i.e. proportional to an identity matrix) we get the classic model, but we can in principle allow it to be correlated and/or heteroscedastic.
It is not too hard to define marginal correlation structures that don't make sense. One class of reasonably sensible models is to always assume an observation-level random effect (as MCMCglmm does for computational reasons) and to allow that random effect to be MVN on the link scale (so that the full model is lognormal-Poisson, logit-normal binomial, etc., depending on the link function and family).
For example, a relatively simple Poisson model with spatially correlated errors might look like this:
$$
\begin{split}
\eta & \sim \textrm{MVN}(a + b x, \Sigma) \\
\Sigma_{ij} & = \sigma^2 \exp(-d_{ij}/s) \\
y_i & \sim \textrm{Poisson}(\lambda=\exp(\eta_i))
\end{split}
$$
That is, the marginal distributions of the response values are Poisson-lognormal, but on the link (log) scale the latent Normal variables underlying the response are *multivariate* normal, with a variance-covariance matrix described by an exponential spatial correlation function with scale parameter $s$.
How can one achieve this?
- These types of models are not implemented in `lme4`, for either LMMs or GLMMs; they are fairly low priority, and it is hard to see how they could be implemented for GLMMs (the equivalent for LMMs is tedious but should be straightforward to implement).
- For LMMs, you can use the spatial/temporal correlation structures that are built into (n)lme
- You can use the spatial/temporal correlation structures available for (n)lme, which include basic geostatistical (space) and ARMA-type (time) models.
```{r findcors,eval=FALSE}
library(sos)
findFn("corStruct")
```
finds additional possibilities in the `ramps` (extended geostatistical) and `ape` (phylogenetic) packages.
- You can use these structures in GLMMs via `MASS::glmmPQL` (see Dormann et al.)
- geepack::geeglm
- geoR, geoRglm (power tools); these are mostly designed for fitting spatial random field GLMMs via MCMC -- not sure that they do random effects other than the spatial random effect
- [R-INLA](http://r-inla.org) (super-power tool)
- it is possible to use AD Model Builder to fit spatial GLMMs, as shown in these [AD Model Builder examples](http://admb-project.org/examples/spatial-models); this capability is not in the `glmmADMB` package (and may not be for a while!), but it would be possible to run AD Model Builder via the R2admb package (requires installing -- and learning! ADMB)
- [geoBUGS](http://mathstat.helsinki.fi/openbugs/Manuals/GeoBUGS/Manual.html), the geostatistical/spatial correlation module for WinBUGS, is another alternative (but again requires going outside of R)
## Penalization/handling complete separation
*Complete separation* occurs in a binary-response model when there is
some linear combination of the parameters that perfectly separates failures
from successes - for example, when all of the observations are zero
for some particular combination of categories. The symptoms of this
problem are unrealistically large parameter estimates; ridiculously
large Wald standard errors (the *Hauck-Donner effect*); and various
warnings.
In particular, binomial `glmer()` models with complete separation can lead to
"Downdated VtV is not positive definite" (e.g. see [here](https://github.com/lme4/lme4/issues/483)) or "PIRLS step-halvings failed to reduce deviance in pwrssUpdate" errors (e.g. see [here](https://github.com/lme4/lme4/issues/179#issuecomment-424445187)). Roughly speaking, the complete separation is likely to appear even if one considers only the fixed effects part of the model (counterarguments or counterexamples welcome!), suggesting two quick-and-dirty diagnostic methods. If `fixed_form` is the formula including only the fixed effects:
- `summary(g1 <- glm(fixed_form, family=binomial, data=...))` will show one or more of the following symptoms:
- warnings that `glm.fit: fitted probabilities numerically 0 or 1 occurred`
- parameter estimates of large magnitude (e.g. `any(abs(g1)>8)`, assuming that predictors are either categorical or scaled to have standard deviations of $\approx 1$)
- extremely large Wald standard errors, and large p-values (*Hauck-Donner effect*)
- the `brglm2` package has a method for detecting complete separation: `library("brglm2"); glm(fixed_form, data = ..., family = binomial, method="detect_separation")`. This should say whether complete separation occurs, and in which (combinations of) variables, e.g.
```
Separation: TRUE
Existence of maximum likelihood estimates
(Intercept) height
Inf Inf
0: finite value, Inf: infinity, -Inf: -infinity
```
If complete separation is occurring between categories of a single categorical fixed-effect predictor with a large number of levels, one option would be to treat this fixed effect as a random effect, which will allow some degree of shrinkage to the mean. (It might be reasonable to specify the variance of this term *a priori* to a large value [minimal shrinkage], rather than trying to estimate it from the data.)
(**TODO**: worked example)
The general approach to handling complete separation in logistic regression
is called *penalized regression*; it's available in the `brglm`,
`brglm2`, `logistf`, and `rms` packages. However, these packages
don't handle mixed models, so the best available *general* approach is to
use a Bayesian method that allows you to set a prior on the fixed effects,
e.g. a Gaussian with standard deviation of 3; this can be done
in any of the Bayesian GLMM packages (e.g. `blme`, `MCMCglmm`, `brms`, ...)
(See [supplementary material for Fox et al. 2016](http://bbolker.github.io/mixedmodels-misc/ecostats_chap.html#digression-complete-separation) for a worked example.)
# Estimation
## What methods are available to fit (estimate) GLMMs?
(adapted from Bolker et al TREE 2009)
```{r glmmtab,echo=FALSE,results="asis"}
methtab <- read.table(sep="|",header=TRUE,text="
Method | Advantages | Disadvantages | Packages
Penalized quasi-likelihood | Flexible, widely implemented | Likelihood inference may be inappropriate; biased for large variance or small means | PROC GLIMMIX (SAS), GLMM (GenStat), glmmPQL (R:MASS), ASREML-R
Laplace approximation | More accurate than PQL | Slower and less flexible than PQL | glmer (R:lme4,lme4a), glmm.admb (R:glmmADMB), INLA, glmmTMB, AD Model Builder, HLM
Gauss-Hermite quadrature | More accurate than Laplace | Slower than Laplace; limited to 2‑3 random effects | PROC NLMIXED (SAS), glmer (R:lme4, lme4a), glmmML (R:glmmML), xtlogit (Stata)
Markov chain Monte Carlo | Highly flexible, arbitrary number of random effects; accurate | Slow, technically challenging, Bayesian framework | MCMCglmm (R:MCMCglmm), rstanarm (R), brms (R), MCMCpack (R), WinBUGS/OpenBUGS (R interface: BRugs/R2WinBUGS), JAGS (R interface: rjags/R2jags), AD Model Builder (R interface: R2admb), glmm.admb (post hoc MCMC after Laplace fit) (R:glmmADMB)"
)
set.alignment(default="left")
pander(methtab,split.table=Inf)
```
## Troubleshooting
- double-check the model specification and the data for mistakes
- center and scale continuous predictor variables (e.g. with `scale()`)
- try all available optimizers (e.g. several different implementations of BOBYQA and Nelder-Mead, L-BFGS-B from `optim`, `nlminb()`, ...). While this will of course be slow for large fits, we consider it the gold standard; if all optimizers converge to values that are practically equivalent (it's up to the user to decide what "practically equivalent means for their case"), then we would consider the model fit to be good enough. For example:
```{r allfit,eval=FALSE}
source(system.file("utils", "allFit.R", package="lme4"))
modelfit.all <- allFit(model)
ss <- summary(modelfit.all)
```
### Convergence warnings
Most of the current advice about troubleshooting `lme4` convergence problems can be found in the help page `?convergence`. That page explains that the convergence tests in the current version of `lme4` (1.1-11, February 2016) generate lots of false positives. We are considering raising the gradient warning threshold to 0.01 in future releases of `lme4`. In addition to the general troubleshooting tips above:
- double-check the Hessian calculation with the more expensive Richardson extrapolation method (see examples)
- restart the fit from the apparent optimum, or from a point perturbed slightly away from the optimum (`getME(model,c("theta","beta"))` should retrieve the parameters in a form suitable to be used as the `start` parameter)
<a id="singular-fit"></a>
<a id="zero-variance"></a>
### Singular models: random effect variances estimated as zero, or correlations estimated as +/- 1
It is very common for overfitted mixed models to result in singular fits. Technically, singularity means that some of the $\boldsymbol \theta$ (variance-covariance Cholesky decomposition) parameters corresponding to diagonal elements of the Cholesky factor are exactly zero, which is the edge of the feasible space, or equivalently that the variance-covariance matrix has some zero eigenvalues (i.e. is positive semidefinite rather than positive definite), or (*almost* equivalently) that some of the variances are estimated as zero or some of the correlations are estimated as +/-1. This commonly occurs in two scenarios:
- small numbers of random-effect levels (e.g. <5), as illustrated in [these simulations](http://rpubs.com/bbolker/4187) and discussed (in a somewhat different, Bayesian context) by @gelman_prior_2006.
- complex random-effects models, e.g. models of the form `(f|g)` where `f` is a categorical variable with a relatively large number of levels, or models with several different random-slopes terms.
- When using `lme4`, singularity is most obviously detectable in the output of `summary.merMod()` or `VarCorr.merMod()` when a variance is estimated as 0 (or very small, i.e. orders of magnitude smaller than other variance components) or when a correlation is estimated as exactly $\pm 1$. However, as pointed out by @bates_parsimonious_2015, singularities in larger variance-covariance matrices can be hard to detect: checking for small values among the diagonal elements of the Cholesky factor is a good start.
```{r sing,eval=FALSE}
theta <- getME(model,"theta")
## diagonal elements are identifiable because they are fitted
## with a lower bound of zero ...
diag.element <- getME(model,"lower")==0
any(theta[diag.element]<1e-5)
```
- In `MCMCglmm`, singular or near-singular models will provoke an error and a requirement to specify a stronger prior.
At present there are a variety of strong opinions about how to resolve such problems. Briefly:
- @barr_random_2013 suggest always starting with the maximal model (i.e. the most random-effects component of the model that is *theoretically* identifiable given the experimental design) and then dropping terms when singularity or non-convergence occurs (please see the paper for detailed recommendations ...)
- @matuschek_balancing_2017 and @bates_parsimonious_2015 strongly disagree, suggesting that models should be simplified *a priori* whenever possible; they also provide [tools](https://github.com/dmbates/RePsychLing) for diagnosing and mitigating singularity.
- One alternative (suggested by Robert LaBudde) for the small-numbers-of-levels scenario is to "fit the model with the random factor as a fixed effect, get the level coefficients in the sum to zero form, and then compute the standard deviation of the coefficients." This is appropriate for users who are (a) primarily interested in measuring variation (i.e. the random effects are not just nuisance parameters, and the variability [rather than the estimated values for each level] is of scientific interest), (b) unable or unwilling to use other approaches (e.g. MCMC with half-Cauchy priors in WinBUGS), (c) unable or unwilling to collect more data. For the simplest case (balanced, orthogonal, nested designs with normal errors) these estimates of standard deviations should equal the classical method-of-moments estimates.
- Bayesian approaches allow the user to specify a informative prior that avoids singularity.
- The `blme` package [@chung_nondegenerate_2013] provides a wrapper for the `lme4` machinery that adds a particular form of weak prior to get an approximate a Bayesian maximum *a posteriori* estimate that avoids singularity.
- The `MCMCglmm` package allows for priors on the variance-covariance matrix
- The `rstanarm` and `brms` packages provide wrappers for the Stan Hamiltonian MCMC engine that fit GLMMs via `lme4` syntax, again allowing a variety of priors to be set.
- If a variance component is zero, dropping it from the model will have no effect on any of the estimated quantities (although it will affect the AIC, as the variance parameter is counted even though it has no effect). @pasch_interspecific_2013 gives one example where random effects were dropped because the variance components were consistently estimated as zero. Conversely, if one chooses for philosophical grounds to retain these parameters, it won't change any of the answers.
### Setting residual variances to a fixed value (zero or other)
For some problems it would be convenient to be able to set the residual variance term to zero, or a fixed value. This is difficult in `lme4`, because the model is parameterized internally in such a way that the residual variance is profiled out (i.e., calculated directly from a residual deviance term) and the random-effects variances are scaled by the residual variance.
[Searching the r-sig-mixed-models list for "fix residual variance"](https://www.google.ca/search?q=site%3A%2F%2Fstat.ethz.ch%2Fpipermail%2Fr-sig-mixed-models%2F+fix+residual+variance)
- This is done in the `metafor` package, for meta-analytic models
- You can use the `blme` package to fix the residual variance: from Vincent Dorie,
```
library(blme)
blmer(formula = y ~ 1 + (1 | group), weights = V,
resid.prior = point(1.0), cov.prior = NULL)
```
This sets the residual variance to 1.0. You *cannot* use this to make it
exactly zero, but you can make it very small (and experiment with setting
it to different small values, e.g. 0.001 vs 0.0001, to see how sensitive
the results are).
- There is an [rrBlupMethod6 package](https://CRAN.R-project.org/package=rrBlupMethod6) on CRAN ("Re-parametrization of mixed model formulation to allow for a fixed residual variance when using RR-BLUP for genom[e]wide estimation of marker effects"), but it seems fairly special-purpose.
- it might be possible *in principle* to adapt `lme4`'s internal `devfun2()` function (used in the likelihood profiling computation for LMMs), which uses a specified value of the residual standard deviation in computing likelihood, but as @bates_fitting_2015 say:
> The resulting function is not useful for general nonlinear optimization — one can easily wander into parameter regimes corresponding to infeasible (non-positive semidefinite) variance-covariance matrices — but it serves for likelihood profiling, where one focal parameter is varied at a time and the optimization over the other parameters is likely to start close to an optimum.
### Other problems/`lme4` error messages
Most of the following error messages are relatively unusual, and happen mostly with complex/large/unstable models. There is often no simple fix; the standard suggestions for troubleshooting are (1) try rescaling and/or centering predictors; (2) see if a simpler model can be made to work; (3) look for severe lack of balance and/or complete separation in the data set.
- `PIRLS step-halvings failed to reduce deviance in pwrssUpdate`
- When using `lme4` to fit GLMMs with link functions that do not automatically constrain the response to the allowable range of the distributional family (e.g. binomial models with a log link, where the estimated probability can be >1, or inverse-Gamma models, where the estimated mean can be negative), it is not unusual to get this error. This occurs because `lme4` doesn't do anything to constrain the predicted values, so `NaN` values pop up, which aren't handled gracefully. If possible, switch to a link function to one that constrains the response (e.g. logit link for binomial or log link for Gamma).
- `Downdated VtV is not positive definite`: no specific advice, see general suggestions above
- `convergence code 3 from bobyqa: bobyqa -- a trust region step failed to reduce q`: again no specific advice about fixing this, although there is a [useful discussion of the meaning of the error message on CrossValidated](https://stats.stackexchange.com/questions/89945/meaning-of-a-convergence-warning-in-glmer)
## REML for GLMMs
- While restricted maximum likelihood (REML) procedures ([Wikipedia](http://en.wikipedia.org/wiki/Restricted_maximum_likelihood) are well established for linear mixed models, it is less clear how one should define and compute the equivalent criteria (integrating out the effects of fixed parameters) for GLMMs. @millar_maximum_2011 and @berger_integrated_1999 are possible starting points in the peer-reviewed literature, and there are mailing-list discussions of these issues [here](https://stat.ethz.ch/pipermail/r-sig-mixed-models/2009q1/002104.html) and [here](http://lists.admb-project.org/pipermail/users/2011-June/001224.html).
- Attempting to use `REML=TRUE` with `glmer` will produce the warning `extra argument(s) ‘REML’ disregarded`
- `glmmTMB` allows `REML=TRUE` for GLMMs (it uses the Laplace approximation to integrate over the fixed effect parameters), since version 0.2.2
# Model diagnostics
# Inference and confidence intervals
## Testing hypotheses
### What are the p-values listed by `summary(glmerfit)` etc.? Are they reliable?
By default, in keeping with the tradition in analysis of generalized linear models, `lme4` and similar packages display the Wald Z-statistics for each parameter in the model summary. These have one big advantage: they're convenient to compute. However, they are asymptotic approximations, assuming both that (1) the sampling distributions of the parameters are multivariate normal (or equivalently that the log-likelihood surface is quadratic) and that (2) the sampling distribution of the log-likelihood is (proportional to) $\chi^2$. The second approximation is discussed further under "Degrees of freedom". The first assumption usually requires an even greater leap of faith, and is known to cause problems in some contexts (for binomial models failures of this assumption are called the *Hauck-Donner effect*), especially with extreme-valued parameters.
### Methods for testing single parameters
From worst to best:
- Wald $Z$-tests
- **For balanced, nested LMMs** where degrees of freedom can be computed according to classical rules: Wald $t$-tests
- Likelihood ratio test, either by setting up the model so that the parameter can be isolated/dropped (via `anova` or `drop1`, or via computing likelihood profiles
- Markov chain Monte Carlo (MCMC) or parametric bootstrap confidence intervals
### Tests of effects (i.e. testing that several parameters are simultaneously zero)
From worst to best:
- Wald chi-square tests (e.g. `car::Anova`)
- Likelihood ratio test (via `anova` or `drop1`)
- **For balanced, nested LMMs** where df can be computed: conditional F-tests
- **For LMMs**: conditional F-tests with df correction (e.g. Kenward-Roger in `pbkrtest` package). (@stroup_rethinking_2014 states (referencing @stroup_non-normal_2013) that K-R actually works reasonably well for GLMMs:
> Notice the non-integer values of the denominator df. They, and the $F$ and $p$ values, reflect the procedure developed by Kenward and Roger (2009) to account for the effect of the covariance structure on degrees of freedom and standard errors. Although the Kenward–Roger adjustment was derived for the LMM with normally distributed data and is an ad hoc procedure for GLMMs with non-normal data, informal simulation studies consistently have suggested that the adjustment is accurate. The Kenward-Roger adjustment requires that the SAS GLIMMIX default computing algorithm, pseudo-likelihood, be used rather than the Laplace algorithm used to obtain AICC statistics. Stroup (2013b) found that for binomial and Poisson GLMMs, pseudo-likelihood with the Kenward–Roger adjustment yields better Type I error control than Laplace while preserving the GLMM’s advantage with respect to power and accuracy in estimating treatment means.
However, K-R is not implemented in R for GLMMs; Stroup suggests that a pseudo-likelihood [@wolfinger_generalized_1993] approach is necessary in order to implement K-R for GLMMs.
- MCMC or parametric, or nonparametric, bootstrap comparisons (nonparametric bootstrapping must be implemented carefully to account for grouping factors)
### Is the likelihood ratio test reliable for mixed models?
- It depends.
- Not for fixed effects in finite-size cases (see @pinheiro_mixed-effects_2000): may depend on 'denominator degrees of freedom' (number of groups) and/or total number of samples - total number of parameters
- Conditional F-tests are preferred for LMMs, **if** denominator degrees of freedom are known
### Why doesn't `lme4` display denominator degrees of freedom/p values? What other options do I have?
There is an [R FAQ entry](http://cran.r-project.org/doc/FAQ/R-FAQ.html#Why-are-p_002dvalues-not-displayed-when-using-lmer_0028_0029_003f) on this topic, which links to a [mailing list post](https://stat.ethz.ch/pipermail/r-help/2006-May/094765.html) by Doug Bates (there is also a voluminous [mailing list thread](http://rwiki.sciviews.org/doku.php?id=guides:lmer-tests) reproduced on the R wiki). The bottom line is
- For special cases that correspond to classical experimental designs (i.e. balanced designs that are nested, split-plot, randomized block, etc.) ... we can show that the null distributions of particular ratios of sums of squares follow an $F$ distribution with known numerator and denominator degrees of freedom (and hence the sampling distributions of particular contrasts are t-distributed with known df). In more complicated situations (unbalanced, GLMMs, crossed random effects, models with temporal or spatial correlation, etc.) it is not in general clear that the null distribution of the computed ratio of sums of squares is really an F distribution, for *any* choice of denominator degrees of freedom.
- For each simple degrees-of-freedom recipe that has been suggested (trace of the hat matrix, etc.) there seems to be at least one fairly simple counterexample where the recipe fails badly (e.g. see [this r-help thread from September 2006](https://stat.ethz.ch/pipermail/r-help/2006-September/112495.html)).
- When the responses are normally distributed and the design is balanced, nested etc. (i.e. the classical LMM situation), the scaled deviances and differences in deviances are exactly $F$-distributed and looking at the experimental design (i.e., which treatments vary/are replicated at which levels) tells us what the relevant degrees of freedom are (see "df alternatives" below)
- Two approaches to approximating df (Satterthwaite and Kenward-Roger) have been implemented in R, Satterthwaite in `lmerTest` and Kenward-Roger in `pbkrtest` (as `KRmodcomp`): although it was derived for LMMs, @stroup_rethinking_2014 states on the basis of (unpresented) simulation results that it actually works reasonably well for GLMMs. However, at present the code in `KRmodcomp` only handles LMMs. (Kenward-Roger is probably the most reliable option [@schaalje_adequacy_2002], although it may be prohibitively computationally expensive for large data sets.)
- Note that there are several different issues at play in finite-size (small-sample) adjustments, which apply slightly differently to LMMs and GLMMs.
- When the data are not classical (crossed, unbalanced, R-side effects), we might still guess that the deviances etc. are approximately F-distributed but that we don't know the real degrees of freedom -- this is what the Satterthwaite, Kenward-Roger, Fai-Cornelius, etc. approximations are supposed to do.
- When the responses are not normally distributed (as in GLMs and GLMMs), and when the scale parameter is not estimated (as in standard Poisson- and binomial-response models), then the deviance differences are only asymptotically F- or chi-square-distributed (i.e. not for our real, finite-size samples). In standard GLM practice, we usually ignore this problem; there is some literature on finite-size corrections for GLMs under the rubrics of "Bartlett corrections" and "higher order asymptotics" (see @McCullaghNelder1989, @cordeiro_improved_1994, @cordeiro_note_1998 and the `cond` package ([on CRAN](https://cran.r-project.org/web/packages/cond/index.html)) [which works with GLMs, not GLMMs]), but it's rarely used. (The bias correction/Firth approach implemented in the `brglm` package attempts to address the problem of finite-size bias, not finite-size non-chi-squaredness of the deviance differences.)
- When the scale parameter in a GLM is estimated rather than fixed (as in Gamma or quasi-likelihood models), it is sometimes recommended to use an $F$ test to account for the uncertainty of the scale parameter (e.g. @venables_modern_2002 recommend `anova(...,test="F")` for quasi-likelihood models)
- Combining these issues, one has to look pretty hard for information on small-sample or finite-size corrections for GLMMs: @feng_small_2004 and @bell_small_2010 look like good starting points, but it's not at all trivial.
#### Df alternatives:
- use MASS::glmmPQL (uses old `nlme` rules approximately equivalent to SAS 'inner-outer'/'within-between' rules) for GLMMs, or `(n)lme` for LMMs
- Guess the denominator df from standard rules (for standard designs, e.g. see @GotelliEllison2004) and apply them to $t$ or $F$ tests
- Run the model in `lme` (if possible) and use the denominator df reported there (which follow a simple 'inner-outer' rule which should correspond to the canonical answer for simple/orthogonal designs), applied to $t$ or $F$ tests. For the explicit specification of the rules that `lme` uses, see page 91 of Pinheiro and Bates (*this page was previously available on [Google Books](http://tinyurl.com/ntygq3), but the link is no longer useful, so here are the relevant paragraphs*):
> These conditional tests for fixed-effects terms require denominator degrees of freedom. In the case of the conditional $F$-tests, the numerator degrees of freedom are also required, being determined by the term itself. The denominator degrees of freedom are determined by the grouping level at which the term is estimated. A term is called inner relative to a factor if its value can change within a given level of the grouping factor. A term is outer to a grouping factor if its value does not changes within levels of the grouping factor. A term is said to be estimated at level $i$, if it is inner to the $i-1$st grouping factor and outer to the $i$th grouping factor. For example, the term `Machine` in the `fm2Machine` model is outer to `Machine %in% Worker` and inner to `Worker`, so it is estimated at level 2 (`Machine %in% Worker`). If a term is inner to all $Q$ grouping factors in a model, it is estimated at the level of the within-group errors, which we denote as the $Q+1$st level.
>
> The intercept, which is the parameter corresponding to the column of all 1's in the model matrices $X_i$, is treated differently from all the other parameters, when it is present. As a parameter it is regarded as being estimated at level 0 because it is outer to all the grouping factors. However, its denominator degrees of freedom are calculated as if it were estimated at level $Q+1$. This is because the intercept is the one parameter that pools information from all the observations at a level even when the corresponding column in $X_i$ doesn't change with the level.
>
> Letting $m_i$ denote the total number of groups in level $i$ (with the convention that $m_0=1$ when the fixed effects model includes an intercept and 0 otherwise, and $m_{Q+1}=N$) and $p_i$ denote the sum of the degrees of freedom corresponding to the terms estimated at level $i$, the $i$th level denominator degrees of freedom is defined as
>
> $$ \mathrm{denDF}_i = m_i - (m_{i-1} + p_i), i = 1, \dots, Q $$
>
> This definition coincides with the classical decomposition of degrees of freedom in balanced, multilevel ANOVA designs and gives a reasonable approximation for more general mixed-effects models.
Note that the implementation used in `lme` **gets the wrong answer for random-slopes models**:
```{r lmeDF,message=FALSE}
library(nlme)
lmeDF <- function(formula=distance~age,random=~1|Subject) {
mod <- lme(formula,random,data=Orthodont)
aa <- anova(mod)
return(setNames(aa[,"denDF"],rownames(aa)))
}
lmeDF()
lmeDF(random=~age|Subject) ## wrong!
```
I (BB) have re-implemented this algorithm in a way that does slightly better for random-slopes models (but may still get confused!), see [here](R/calcDenDF.R).
```{r calcDenDF}
source("R/calcDenDF.R")
calcDenDF(~age,"Subject",nlme::Orthodont)
calcDenDF(~age,data=nlme::Orthodont,random=~1|Subject)
calcDenDF(~age,data=nlme::Orthodont,random=~age|Subject) ## off by 1
```
- use SAS, Genstat (AS-REML), Stata?
- Assume infinite denominator df (i.e. $Z$/$\chi^2$ test rather than $t$/$F$) if number of groups is large (>45? Various rules of thumb for how large is "approximately infinite" have been posed, including [in @angrist_mostly_2009], 42 (in homage to Douglas Adams)
### Testing significance of random effects
- the most common way to do this is to use a likelihood ratio test, i.e. fit the full and reduced models (the reduced model is the model with the focal variance(s) set to zero). For example:
```{r raneff_test,message=FALSE,cache=TRUE}
library(lme4)
m2 <- lmer(Reaction~Days+(1|Subject)+(0+Days|Subject),sleepstudy,REML=FALSE)
m1 <- update(m2,.~Days+(1|Subject))
m0 <- lm(Reaction~Days,sleepstudy)
anova(m2,m1,m0) ## two sequential tests
```
With recent versions of `lme4`, goodness-of-fit (deviance) can be compared between `(g)lmer` and `(g)lm` models, although `anova()` must be called with the mixed (`(g)lmer`) model listed first.
Keep in mind that LRT-based null hypothesis tests are conservative when the null value (such as $\sigma^2=0$) is on the boundary of the feasible space [@self_asymptotic_1987;@stram_variance_1994;@GoldmanWhelan2000]; in the simplest case (single random effect variance), the p-value is approximately twice as large as it should be [@pinheiro_mixed-effects_2000].
- Consider *not* testing the significance of random effects. If the random effect is part of the experimental design, this procedure may be considered 'sacrificial pseudoreplication' [@Hurlbert1984]. Using stepwise approaches to eliminate non-significant terms in order to squeeze more significance out of the remaining terms is dangerous in any case.
- consider using the `RLRsim` package, which has a fast implementation of simulation-based tests of null hypotheses about zero variances, for simple tests. (However, it only applies to `lmer` models, and is a bit tricky to use for more complex models.)
```{r RLRsim,message=FALSE}
library(RLRsim)
## compare m0 and m1
exactLRT(m1,m0)
## compare m1 and m2
mA <- update(m2,REML=TRUE)
m0 <- update(mA, . ~ . - (0 + Days|Subject))
m.slope <- update(mA, . ~ . - (1|Subject))
exactRLRT(m0=m0,m=m.slope,mA=mA)
```
- Parametric bootstrap: fit the reduced model, then repeatedly simulate from it and compute the differences between the deviance of the reduced and the full model for each simulated data set. Compare this null distribution to the observed deviance difference. This procedure is implemented in the `pbkrtest` package.
```{r pboot_vartest_comp,cache=TRUE}
(pb <- pbkrtest::PBmodcomp(m2,m1,seed=101))
```
### Standard errors of variance estimates
<a id="variance-standard-errors"></a>
- Paraphrasing Doug Bates: the sampling distribution of variance estimates is in general strongly asymmetric: the standard error may be a poor characterization of the uncertainty.
- `lme4` allows for computing likelihood profiles of variances and computing confidence intervals on their basis; these likelihood profile confidence intervals are subject to the usual caveats about the LRT with finite sample sizes.
- Using an MCMC-based approach (the simplest/most canned is probably to use the `MCMCglmm` package, although its mode specifications are not identical to those of lme4) will provide posterior distributions of the variance parameters: quantiles or credible intervals (`HPDinterval()` in the `coda` package) will characterize the uncertainty.
- (don't say we didn't warn you ...) `[n]lme` fits contain an element called `apVar` which contains the approximate variance-covariance matrix (derived from the Hessian, the matrix of (numerically approximated) second derivatives of the likelihood (REML?) at the maximum (restricted?) likelihood values): you can derive the standard errors from this list element via `sqrt(diag(lme.obj$apVar))`. For whatever it's worth, though, [these estimates might not match](http://www.biostat.wustl.edu/archives/html/s-news/2003-07/msg00127.html) the [estimates that SAS gives](http://www.tau.ac.il/cc/pages/docs/sas8/stat/chap41/sect25.htm#mixedcpe) which are supposedly derived in the same way.
- it's not a full solution, but there is some more information [here](https://rpubs.com/bbolker/waldvar). I have some delta-method computations there that are off by a factor of 2 for the residual standard deviation, as well as some computations based on reparameterizing the deviance function.
### P-values: MCMC and parametric bootstrap
Abandoning the approximate $F$/$t$-statistic route, one ends up with the more general problem of estimating $p$-values. There is a wider range of options here, although many of them are computationally intensive ...
#### Markov chain Monte Carlo sampling:
- pseudo-Bayesian: post-hoc sampling, typically (1) assuming flat priors and (2) starting from the MLE, possibly using the approximate variance-covariance estimate to choose a candidate distribution
- via `mcmcsamp` (if available for your problem: i.e. LMMs with simple random effects -- not GLMMs or complex random effects)
- via `pvals.fnc` in the `languageR` package, a wrapper for mcmcsamp)
- in AD Model Builder, possibly via the `glmmADMB` package (use the `mcmc=TRUE` option) or the `R2admb` package (write your own model definition in AD Model Builder), or outside of R
- via the `sim` function from the `arm` package (simulates the posterior only for the beta (fixed-effect) coefficients; not yet working with development lme4; would like a better formal description of the algorithm ...?)
- fully Bayesian approaches
- via the `MCMCglmm` package
- `glmmBUGS` (a WinBUGS wrapper/R interface)
- JAGS/WinBUGS/OpenBUGS etc., via the `rjags`/`r2jags`/`R2WinBUGS`/`BRugs` packages
#### Status of mcmcsamp
`mcmcsamp` is a function for lme4 that is supposed to sample from the posterior distribution of the parameters, based on flat/improper priors for the parameters [ed: I believe, but am not sure, that these priors are flat **on the scale of the theta (Cholesky-factor) parameters**]. At present, in the CRAN version (lme4 0.999999-0) and the R-forge "stable" version (lme4.0 0.999999-1), this covers only linear mixed models with uncorrelated random effects.
As has been discussed in a variety of places (e.g. [on r-sig-mixed models](http://article.gmane.org/gmane.comp.lang.r.lme4.devel/1788/), and [on the r-forge bug tracker](https://r-forge.r-project.org/tracker/?func=detail&aid=68&group_id=60&atid=298), it is challenging to come up with a sampler that accounts properly for the possibility that the posterior distributions for some of the variance components may be mixtures of point masses at zero and continuous distributions. Naive samplers are likely to get stuck at or near zero. Doug Bates has always been a bit unsure that `mcmcsamp` is really performing as intended, even in the limited cases it now handles.
Given this uncertainty about how even the basic version works, the `lme4` developers have been reluctant to make the effort to extend it to GLMMs or more complex LMMs, or to implement it for the development version of lme4 ... so unless something miraculous happens, it will not be implemented for the new version of `lme4`. As always, users are encouraged to write and share their own code that implements these capabilities ...
#### Parametric bootstrap
The idea here is that in order to do inference on the effect of (a) predictor(s), you (1) fit the reduced model (without the predictors) to the data; (2) many times, (2a) simulate data from the reduced model; (2b) fit both the reduced and the full model to the simulated (null) data; (2c) compute some statistic(s) [e.g. t-statistic of the focal parameter, or the log-likelihood or deviance difference between the models]; (3) compare the observed values of the statistic from fitting your full model to the data to the null distribution generated in step 2.
- `PBmodcomp` in the `pbkrtest` package
- see the example in `help("simulate-mer")` in the `lme4` package to roll your own, using a combination of `simulate()` and `refit()`.
- `bootMer` in `lme4` version >1.0.0
- a presentation at UseR! 2009 ([abstract](http://www.agrocampus-ouest.fr/math/useR-2009/abstracts/pdf/SanchezEspigares+Ocana.pdf), [slides](http://www.agrocampus-ouest.fr/math/useR-2009/slides/SanchezEspigares+Ocana.pdf)) went into detail about a proposed `bootMer` package and suggested it could work for GLMMs too --
but it does not seem to be active.
## Predictions and/or confidence (or prediction) intervals on predictions
Note that none of the following approaches takes the uncertainty
of the random effects parameters into account ...
The general recipe for computing predictions from a linear or generalized linear model is to
- figure out the model matrix $X$ corresponding to the new data;
- matrix-multiply $X$ by the parameter vector $\beta$ to get the predictions (or linear predictor in the case of GLM(M)s);
- extract the variance-covariance matrix of the parameters $V$
- compute $X V X^{\prime}$ to get the variance-covariance matrix of the predictions;
- extract the diagonal of this matrix to get variances of predictions;
- if computing prediction rather than confidence intervals, add the residual variance;
- take the square-root of the variances to get the standard deviations (errors) of the predictions;
- compute confidence intervals based on a Normal approximation;
- for GL(M)Ms, run the confidence interval boundaries (not the standard errors) through the inverse-link function.
### lme
```{r lmepred,cache=TRUE,message=FALSE}
library(nlme)
fm1 <- lme(distance ~ age*Sex, random = ~ 1 + age | Subject,
data = Orthodont)
plot(Orthodont,asp="fill") ## plot responses by individual
## note that expand.grid() orders factor levels by *order of
## appearance* -- must match levels(Orthodont$Sex)
newdat <- expand.grid(age=c(8,10,12,14), Sex=c("Female","Male"))
newdat$pred <- predict(fm1, newdat, level = 0)
## [-2] drops response from formula
Designmat <- model.matrix(formula(fm1)[-2], newdat)
predvar <- diag(Designmat %*% vcov(fm1) %*% t(Designmat))
newdat$SE <- sqrt(predvar)
newdat$SE2 <- sqrt(predvar+fm1$sigma^2)
library(ggplot2)
pd <- position_dodge(width=0.4)
g0 <- ggplot(newdat,aes(x=age,y=pred,colour=Sex))+
geom_point(position=pd)
cmult <- 2 ## could use 1.96 instead
g0 + geom_linerange(aes(ymin=pred-cmult*SE,ymax=pred+cmult*SE), position=pd)
## prediction intervals
g0 + geom_linerange(aes(ymin=pred-cmult*SE2,ymax=pred+cmult*SE2), position=pd)
```
A similar answer is laid out in the responses to this [StackOverflow question](http://stackoverflow.com/questions/14358811/extract-prediction-band-from-lme-fit).
### lme4
Current versions of lme4 have a `predict` method.
```{r lme4pred,cache=TRUE}
library(lme4)
library(ggplot2)
data("Orthodont",package="MEMSS")
fm1 <- lmer(
formula = distance ~ age*Sex + (age|Subject)
, data = Orthodont
)
newdat <- expand.grid(
age=c(8,10,12,14)
, Sex=c("Female","Male")
, distance = 0
)
newdat$distance <- predict(fm1,newdat,re.form=NA)
mm <- model.matrix(terms(fm1),newdat)
## or newdat$distance <- mm %*% fixef(fm1)
pvar1 <- diag(mm %*% tcrossprod(vcov(fm1),mm))
tvar1 <- pvar1+VarCorr(fm1)$Subject[1] ## must be adapted for more complex models
cmult <- 2 ## could use 1.96
newdat <- data.frame(
newdat
, plo = newdat$distance-cmult*sqrt(pvar1)
, phi = newdat$distance+cmult*sqrt(pvar1)
, tlo = newdat$distance-cmult*sqrt(tvar1)
, thi = newdat$distance+cmult*sqrt(tvar1)
)
#plot confidence
g0 <- ggplot(newdat, aes(x=age, y=distance, colour=Sex))+geom_point()
g0 + geom_pointrange(aes(ymin = plo, ymax = phi))+
labs(title="CI based on fixed-effects uncertainty ONLY")
#plot prediction
g0 + geom_pointrange(aes(ymin = tlo, ymax = thi))+
labs(title="CI based on FE uncertainty + RE variance")
rm("Orthodont") ## clean up
```
### glmmadmb
```{r glmmadmbpred,cache=TRUE}
library(glmmADMB)
data(Orthodont,package="nlme")
fm2 <- glmmadmb(distance ~ age*Sex, random = ~ 1 + age | Subject,
data = Orthodont,
family="gaussian")
## make prediction data frame
newdat <- expand.grid(age=c(8,10,12,14), Sex=c("Female","Male"))
## design matrix (fixed effects)
mm <- model.matrix(delete.response(terms(fm2)),newdat)
## linear predictor (for GLMMs, back-transform this with the
## inverse link function (e.g. plogis() for binomial, beta;
## exp() for Poisson, negative binomial
newdat$distance <- drop(mm %*% fixef(fm2))
predvar <- diag(mm %*% vcov(fm2) %*% t(mm))
newdat$SE <- sqrt(predvar)
newdat$SE2 <- sqrt(predvar+fm2$alpha^2)
```
(Probably overly complicated) `ggplot` code:
```{r ggplot}
library(ggplot2); theme_set(theme_bw())
pd <- position_dodge(width=0.4)
g0 <- ggplot(Orthodont,aes(x=age,y=distance,colour=Sex))+
stat_sum(alpha=0.2,aes(size=..n..))+
scale_size_continuous(breaks=1:4,range=c(2,5))
g1 <- g0+geom_line(data=newdat,position=pd)+
geom_point(data=newdat,shape=17,size=3,position=pd)
g1 + geom_linerange(data=newdat,
aes(ymin=distance-2*SE,ymax=distance+2*SE), position=pd)
## prediction intervals
g1 + geom_linerange(data=newdat,
aes(ymin=distance-2*SE2,ymax=distance+2*SE2), position=pd)
```
## Confidence intervals on conditional means/BLUPs/random effects
### lme4
(From Harold Doran, updated by Assaf Oron Nov. 2013:)
If you want the standard errors of the conditional means, then you can extract them as follows:
```{r stder_condmeans}
library(lme4)
fm1 <- lmer(Reaction ~ Days + (Days|Subject), sleepstudy)
cV <- ranef(fm1, condVar = TRUE)
```
`cV` is a list; each element is a data frame containing the conditional modes for a particular grouping factor. If you use scalar random effects (typically random intercepts), then specifying `ranef(...,drop=TRUE)` will return the conditional modes as a single named vector instead.
The conditional variances are returned as an attribute of the conditional modes.
For example, if we set
```{r ranvar}
ranvar <- attr(cV[[1]], "postVar")
```
then `ranvar` is a 3-D array (the attribute is still called `postVar`, rather than `condVar`, for historical reasons/backward compatibility). Individual-level covariance matrices of the conditional modes will sit on the `[,,i]` faces. For example, `ranvar[,,1]` is the variance-covariance matrix of the conditional distribution for the first group, so
```{r sqrtdiagranvars}
sqrt(diag(ranvar[,,1]))
```
will provide the intercept and slope standard standard deviations for the first group's conditional modes. If you have a scalar random effect and used `drop=TRUE` in `ranef()`, then you will (mercifully) receive only a vector of individual variances here (one for each level of the grouping factor). The following incantation will give a matrix of conditional variances with one row for each group and one column for each parameters:
```{r ranse}
ng <- dim(ranvar)[3]
np <- dim(ranvar)[2]
matrix(ranvar[cbind(rep(seq(np),ng),
rep(seq(np),ng),
rep(ng,each=np))],
byrow=TRUE,
nrow=ng)
```
It may be more convenient to use `as.data.frame(cV)` to extract all of the conditional means and standard deviations.
See also section 1.6 of FIXME Batesdraft
Also, note that at a minimum, the uncertainty of the fixed-effect intercept estimate needs to be incorporated as well; probably manually (remember, the model assumes the fixed and random effects to be orthogonal, so this is straightforward). If you predict on an individual with specific fixed-effect covariate values, those need incorporation too.
## Power analysis
Power analysis for (G)LMMs is mostly done by simulation, although
there are some closed-form solutions and approximations, e.g.
@snijders_standard_1993 (Snijders has links to programs and other resources
on [his web page](https://www.stats.ox.ac.uk/~snijders/multilevel.htm)). There are resources and bits of code examples spread all over the internet. e.g. [here](https://rpubs.com/bbolker/11703).
@kain_practical_2015 and @johnson_power_2015 are peer-reviewed papers
that discuss power analysis via simulation in more detail.
```{r eval=FALSE}
library(sos); findFn("{power analysis} mixed simulation")
```
locates the `fullfact`, `pamm`, `simr`, and `simglm` packages.
Depending on the goal, one of these packages may have sufficient
flexibility to do what you want.
# Model selection and averaging
## Can I use AIC for mixed models? How do I count the number of degrees of freedom for a random effect?
- Yes, with caution.
- It depends on the "level of focus" (*sensu* @spiegelhalter_bayesian_2002) whether (e.g.) a single random-effect variance should be counted as 1 degree of freedom (i.e., the variance parameter or as a value between 1 and N-1 (where N is the number of groups): see @vaida_conditional_2005 and @greven_behaviour_2010. If you are interested in population-level prediction/inference, then the former (called *marginal AIC* [mAIC]); if individual-level prediction/inference (i.e., using the BLUPs/conditional modes), then the latter (called *conditional AIC* [cAIC]). Greven and Kneib present results for linear models, giving a version of cAIC that is both computationally efficient and takes uncertainty in the estimation of the variances into account. (Bob O'Hara has a very nice, illustrative [blog post](http://deepthoughtsandsilliness.blogspot.ca/2007/12/focus-on-dic.html) on this topic in the context of DIC ...)
- in cases when testing a variance parameter, AIC may be subject to the same kinds of boundary effects as likelihood ratio test p-values (i.e., AICs may be conservative/overfit slightly when the nested parameter value is on the boundary of the feasible space). @greven_behaviour_2010 explore the problems with mAIC in this context, but do not suggest a solution (they point out that @hughes_model_2003 present a 'one-sided' AIC, but not one that deals with the non-independence of data points. I haven't read Hughes and King, I should go do that).
- AIC also inherits the primary problem of likelihood ratio tests in the GLMM context -- that is, that LRTs are asymptotic tests. A finite-size correction for AIC does exist (AICc) -- but it was developed in the context of linear models. As far as I know its adequacy in the GLMM case has not been established. See e.g. @richards_testing_2005 for caution about AICc in the GLM (not GLMM) case.
- lme4 and nlme count parameters for AIC(c) as follows:
- the number of fixed-effect parameters is straightforward (the length of the fixed-effect parameter vector beta, i.e. `length(fixef(model))`)
- each random term in the model with $q$ components counts for $q(q+1)/2$ parameters -- for example, a term of the form (slope|group) has 3 parameters (intercept variance, slope variance, correlation between intercept and slope).
- models that use a scale parameter (e.g. the variance parameter of linear mixed models, or the scale parameter of a Gamma GLMM -- standard GLMMs such as binomial and Poisson do not) get an extra parameter counted. Whether to add nuisance parameters or not, such as the residual variance parameter (estimated based on the residual variance, rather than an explicit parameter in the optimization) is as far as I know an open question. In the classic AIC context it doesn't matter as long as one is consistent. In the AICc context, I don't think anyone really knows the answer ... adding +1 for the residual variance parameter (as lme4 does) would make the model selection process slightly more conservative.
# Model summaries (goodness-of-fit, decomposition of variance, etc.)
### How do I compute a coefficient of determination ($R^2$), or an analogue, for (G)LMMs?
#### Problem
(*This topic applies to both LMMs and GLMMs, perhaps more so to LMMs, because the issues are even harder for GLMMs.*)
Researchers often want to know if there is a simple (or at least implemented-in-R) way to get an analogue of $R^2$ or another simple goodness-of-fit metric for LMMs or GLMMs. This is a challenging question in both the GLM and LMM worlds (and therefore doubly so for GLMMs), because it turns out that the wonderful simplicity of $R^2$ breaks down in the extension to GLMs or LMMs. If you're trying to quantify "fraction of variance explained" in the GLM context, should you include or exclude sampling variation (e.g., Poisson variation around the expected mean)? [According to an `sos::findFn` search for "Nagelkerke", one of the common solutions to this problem, the `LogRegR2` function in the `descr` package computes several different "pseudo-$R^2$" measures for logistic regression.] If you're trying to quantify it in the LMM context, should you include or exclude variation of different random-effects terms?
The same questions apply more generally to decomposition of variance (i.e. trying to assess the contribution of various model components to the overall fit, not just trying to assess the overall goodness-of-fit of the model); there is unlikely to be a single recipe that does everything you want.
This has been discussed at various times on the mailing lists. [This thread](http://thread.gmane.org/gmane.comp.lang.r.lme4.devel/3281) and [this thread](http://thread.gmane.org/gmane.comp.lang.r.lme4.devel/684) on the r-sig-mixed-models mailing list are good starting points, and [this post](http://thread.gmane.org/gmane.comp.lang.r.lme4.devel/2143) is useful too.
In one of those threads, Doug Bates said:
> Assuming that one wants to define an R^2 measure, I think an
> argument could be made for treating the penalized residual sum of
> squares from a linear mixed model in the same way that we consider the
> residual sum of squares from a linear model. Or one could use just
> the residual sum of squares without the penalty or the minimum
> residual sum of squares obtainable from a given set of terms, which
> corresponds to an infinite precision matrix. I don't know, really.
> It depends on what you are trying to characterize.
#### Simple/crude solutions
In one of those threads, Jarrett Byrnes contributed the following code:
```{r r2corr}
r2.corr.mer <- function(m) {
lmfit <- lm(model.response(model.frame(m)) ~ fitted(m))
summary(lmfit)$r.squared
}
```
$\Omega^2_0$ [@xu_measuring_2003], which is almost the same, is based on comparing the residual variance of the full model against the residual variance of a (fixed) intercept-only null model:
```{r resvar,eval=FALSE}
1-var(residuals(m))/(var(model.response(model.frame(m)))
```
#### Sophisticated solutions
@gelman_bayesian_2006 propose/discuss Bayesian measures of $R^2$ (I don't know if anyone has created a canned implementation of these measures in R).
@nakagawa_general_2013 and @johnson_extension_2014 have also proposed a general methodology for computing $R^2$; J. Lefcheck gives examples [here](https://jonlefcheck.net/2013/03/13/r2-for-linear-mixed-effects-models/) and [here](https://github.com/jslefche/piecewiseSEM/blob/master/README.md#get-r2-for-individual-models), based on his implementation in the `piecewiseSEM` package ([CRAN](https://cran.r-project.org/web/packages/piecewiseSEM/index.html), [Github](https://github.com/jslefche/piecewiseSEM)). See also @rights_quantifying_2018 ...
A related question is how to quantify "repeatability" (i.e., ratios of variance at different levels) in GLMMs, especially how to compute the "residual error" term for GLMMs: see @nakagawa_repeatability_2010 and the [rptR package](http://rptr.r-forge.r-project.org/).
The bottom line is that there are some simple recipes (and some more complex recipes that may or may not have been coded up by someone), but that '''you have to think carefully about what information you want to get out of the coefficient of determination''', because no recipe will have all of the properties of $R^2$ in the simple linear model case.
**Packages/functions**: `sjstats::r2`, `MuMin::r.squaredGLMM`, `r2glmm` package, the standalone [r2MLM function](https://my.vanderbilt.edu/jasonrights/software/r2MLM), stuff in the `piecewiseSEM` package ...
### Do I have to specify the levels of fixed effects in lmer?
No. See Doug Bates reply to this question [here](http://r.789695.n4.nabble.com/lme4-and-Variable-level-detection-td881680.html)
# Miscellaneous/procedural
## Pronunciation of `lmer`/`glmer`/etc.
- `lmer`: I have heard "ell emm ee arr" (i.e. pronouncing each letter); "elmer" (probably most common); and "lemur"
- `glmer`: "gee ell emm ee arr", "gee elmer", "glimmer", or "gleamer"
- for `lme` and `nlme` people just seem to spell out the names (rather than saying e.g. "lemmy" and "nelmy")
## Storing information
Recent versions of `lme4` output contain an `@optinfo` slot that stores warnings.
Copied from https://stat.ethz.ch/pipermail/r-help/2012-February/302767.html :
> There's a somewhat hack-ish solution, which is to use options(warn=2) to 'upgrade' warnings to errors, and then use try() or tryCatch() to catch them.
> More fancily, I used code that looked something like this to save warnings as I went along (sorry about the <<- ) in a recent simulation study. You could also check w$message to do different things in the case of different warnings.
```{r catch_errors,eval=FALSE}
## n.b. have to set up a 3D warn array first ...
withCallingHandlers(tryCatch(fun(n=nvec[j],tau=tauvec[i],...),
error = function(e) {
warn[k,i,j] <<- paste("ERROR:",e$message)
NA_ans}),
warning = function(w) {
warn[k,i,j] <<- w$message
invokeRestart("muffleWarning")
})
```
# Mixed modeling packages
- also see the [package comparison](http://glmm.wikidot.com/pkg-comparison) on `glmm.wikidot.com`
## Which R packages (functions) fit GLMMs?
- MASS::glmmPQL (penalized quasi-likelihood)
- lme4::glmer (Laplace approximation and adaptive Gauss-Hermite quadrature [AGHQ])
- MCMCglmm (Markov chain Monte Carlo)
- glmmML (AGHQ)
- glmmAK (AGHQ?)
- glmmADMB (Laplace)
- glmm (from Jim Lindsey's `repeated` package: AGHQ)
- gamlss.mx
- ASREML-R
- sabreR
## Should I use `aov()`, `nlme`, or `lme4`, or some other package?
- `aov()` (in the `stats` package in base R: balanced, orthogonal designs only (R analogue of SAS PROC GLM)
- `nlme` (analogue of SAS `PROC MIXED`/`NLMIXED`)
- allows more complex designs than `aov` (unbalanced, heteroscedasticity and/or correlation among residual errors)
- more mature than `lme4`
- well-documented [@pinheiro_mixed-effects_2000]
- implements R-side effects (heteroscedasticity and correlation)
- estimates "denominator degrees of freedom" for $F$ statistics, and hence $p$ values, for LMMs (but see above)
- `lme4` (also SAS `PROC MIXED`/`NLMIXED`):
- fastest
- best for crossed designs (although they are possible in `lme`)
- GLMMs
- cutting-edge (for better or worse!)
- likelihood profiles
- use `lme4` for GLMMs, or if you have big data (thousands to tens of thousands of records)
The following is modified from a contribution by Kingsford Jones, found 2010-03-16
### linear and nonlinear mixed models
- [lme](http://cran.r-project.org/web/packages/lme4/index.html) -- Linear mixed-effects models using S4 classes
- `lmm` -- Linear mixed models
- `nlme` -- Linear and Nonlinear Mixed Effects Models
- `sabreR`
- [regress](http://cran.r-project.org/web/packages/regress/index.html) Linear mixed models
### GLMMs
- glmmAK -- Generalized Linear Mixed Models
- MASS -- Main Package of Venables and Ripley's MASS (see function glmmPQL)
- MCMCglmm -- MCMC Generalised Linear Mixed Models
- lme4 (glmer)
- glmmML
- gamlss.mx
- sabreR
### Additive and generalized-additive mixed models
- amer -- Additive mixed models with lme4
- gamm4 -- Generalized additive mixed models using mgcv and lme4
- mgcv (gamm function, via glmmPQL in MASS package)
- gamlss.mx
### Hierarchical GLMs
- hglm -- hglm is used to fit hierarchical generalized linear models
- HGLMMM -- Hierarchical Generalized Linear Models
### diagnostic and modeling frameworks
- influence.ME -- Tools for detecting influential data in mixed effects models
- arm -- Data Analysis Using Regression and Multilevel/Hierarchical Models
- pamm -- Power analysis for random effects in mixed models
- RLRsim -- Exact (Restricted) Likelihood Ratio tests for mixed and additive models
- npde -- Normalised prediction distribution errors for nonlinear mixed-effect models
- multilevel -- Multilevel Functions (psychology-oriented; within-group agreement, random group resampling, etc.)
- languageR
- pbkrtest -- parametric bootstrap and Kenward-Roger tests
### data and examples
- MEMSS -- Data sets from Mixed-effects Models in S
- mlmRev -- Examples from Multilevel Modelling Software Review
- SASmixed -- Data sets from "SAS System for Mixed Models"
### extensions
- lmeSplines -- lmeSplines
- lmec -- Linear Mixed-Effects Models with Censored Responses
- kinship -- mixed-effects Cox models, sparse matrices, and modeling data from large pedigrees
- coxme -- Mixed Effects Cox Models
- ordinal -- Regression Models for Ordinal Data
- phmm -- Proportional Hazards Mixed-effects Model (PHMM)
- pedigreemm -- Pedigree-based mixed-effects models
- (see also MCMCglmm for pedigree-based approaches)
- heavy -- Estimation in the linear mixed model using heavy-tailed distributions
- GLMMarp -- Generalized Linear Multilevel Model with AR(p) Errors Package
- glmmlasso -- penalized GLMM fitting
- spatialCovariance -- spatial covariance matrix calculations
### Interfaces to other systems
- glmmBUGS -- Generalised Linear Mixed Models and Spatial Models with BUGS
- Interfaces to WinBUGS/OpenBUGS/JAGS (roll your own model file):
* R2WinBUGS
* r2jags
* rjags
* RBugs
### modeling based on LMMs
- `nlmeODE` -- Non-linear mixed-effects modelling in nlme using differential equations
- `longRPart` -- Recursive partitioning of longitudinal data using mixed-effects models
- `PSM` -- Non-Linear Mixed-Effects modelling using Stochastic Differential Equations
## Off-CRAN mixed modeling packages:
### R-forge and Github:
- `glmmADMB` (R-forge, interface to AD Model Builder)
- `spida`, `p3d` (Georges Monette)
### Other open source:
- [bernor](http://www.stat.umn.edu/geyer/bernor/ bernor) package (logit-normal fitting), by Yun Ju Sung and Charles J. Geyer
- `glmm` (in Jim Lindsey's `repeated` package: at [Lindsey's web site](http://www.commanster.eu/rcode.html)
### Commercial:
- `OpenMx` -- Advanced Structural Equation Modeling
- `ASReml-R` (commercial, but 30 days' free use/free license for academic or developing-country use available). Very good at complex LMMs (fast, flexible covariance structures, etc.), but only offers PQL for GLMMs, and the manual says:
> we cannot recommend the use of this technique for general use. It is included in the current version of asreml() for advanced users. It is highly recommended that its use be accompanied by some form of cross-validatory assessment for the specific dataset concerned."
Resources:
- [short R wiki tutorial](http://rwiki.sciviews.org/doku.php?id=guides:tutorials:asreml)
- [reference manual](http://www.vsni.co.uk/downloads/asreml/release3/asreml-R.pdf reference manual) (PDF)
- Luis Apiolaza's [asreml-r cookbook](http://apiolaza.net/asreml-r/)
## Package versions used
```{r sessioninfo}
sessionInfo()
```
## To do
- update package descriptions; cross-link with [Task View](http://bbolker.github.io/mixedmodels-misc/MixedModels.html) ? `rethinking`, `brms`, ...
- more on post-analysis (`broom(.mixed)`, `emmeans`, `multcomp`, ...)
- more on confidence intervals, simulating from conditional distributions, etc.)
# Bibliography