Package 'RLRsim' reference manual

Title:	Exact (Restricted) Likelihood Ratio Tests for Mixed and Additive Models
Description:	Rapid, simulation-based exact (restricted) likelihood ratio tests for testing the presence of variance components/nonparametric terms for models fit with nlme::lme(),lme4::lmer(), lmeTest::lmer(), gamm4::gamm4(), mgcv::gamm() and SemiPar::spm().
Authors:	Fabian Scheipl [aut, cre] , Ben Bolker [ctb]
Maintainer:	Fabian Scheipl <[email protected]>
License:	GPL
Version:	3.1-8
Built:	2024-11-04 05:03:58 UTC
Source:	https://github.com/fabian-s/rlrsim

R package for fast and exact (restricted) likelihood ratio tests for mixed and additive models.

Description

RLRsim implements fast simulation-based exact tests for variance components in mixed and additive models for conditionally Gaussian responses – i.e., tests for questions like:

is the variance of my random intercept significantly different from 0?
is this smooth effect significantly nonlinear?
is this smooth effect significantly different from a constant effect?

The convenience functions exactRLRT and exactLRT can deal with fitted models from packages lme4, nlme, gamm4, SemiPar and from mgcv's gamm()-function. Workhorse functions LRTSim and RLRTSim accept design matrices as inputs directly and can thus be used more generally to generate exact critical values for the corresponding (restricted) likelihood ratio tests.

The theory behind these tests was first developed in:
Crainiceanu, C. and Ruppert, D. (2004) Likelihood ratio tests in linear mixed models with one variance component, Journal of the Royal Statistical Society: Series B, 66, 165–185.

Power analyses and sensitivity studies for RLRsim can be found in:
Scheipl, F., Greven, S. and Kuechenhoff, H. (2008) Size and power of tests for a zero random effect variance or polynomial regression in additive and linear mixed models. Computational Statistics and Data Analysis, 52(7), 3283–3299, doi:10.1016/j.csda.2007.10.022.

Author(s)

Fabian Scheipl ([email protected]), Ben Bolker

Likelihood Ratio Tests for simple linear mixed models

Description

This function provides an exact likelihood ratio test based on simulated values from the finite sample distribution for simultaneous testing of the presence of the variance component and some restrictions of the fixed effects in a simple linear mixed model with known correlation structure of the random effect and i.i.d. errors.

Usage

exactLRT(
  m,
  m0,
  seed = NA,
  nsim = 10000,
  log.grid.hi = 8,
  log.grid.lo = -10,
  gridlength = 200,
  parallel = c("no", "multicore", "snow"),
  ncpus = 1L,
  cl = NULL
)
exactLRT(
  m,
  m0,
  seed = NA,
  nsim = 10000,
  log.grid.hi = 8,
  log.grid.lo = -10,
  gridlength = 200,
  parallel = c("no", "multicore", "snow"),
  ncpus = 1L,
  cl = NULL
)

Arguments

`m`	The fitted model under the alternative; of class `lme`, `lmerMod` or `spm`
`m0`	The fitted model under the null hypothesis; of class `lm`
`seed`	Specify a seed for `set.seed`
`nsim`	Number of values to simulate
`log.grid.hi`	Lower value of the grid on the log scale. See `exactLRT`.
`log.grid.lo`	Lower value of the grid on the log scale. See `exactLRT`.
`gridlength`	Length of the grid. See `LRTSim`.
`parallel`	The type of parallel operation to be used (if any). If missing, the default is "no parallelization").
`ncpus`	integer: number of processes to be used in parallel operation: typically one would chose this to the number of available CPUs. Defaults to 1, i.e., no parallelization.
`cl`	An optional parallel or snow cluster for use if parallel = "snow". If not supplied, a cluster on the local machine is created for the duration of the call.

Details

The model under the alternative must be a linear mixed model $y=X\beta+Zb+\varepsilon$ with a single random effect $b$ with known correlation structure and error terms that are i.i.d. The hypothesis to be tested must be of the form

$H_0: \beta_{p+1-q}=\beta^0_{p+1-q},\dots,\beta_{p}=\beta^0_{p};\quad$

$Var(b)=0$

versus

$H_A:\; \beta_{p+1-q}\neq \beta^0_{p+1-q}\;\mbox{or}\dots$

$\mbox{or}\;\beta_{p}\neq \beta^0_{p}\;\;\mbox{or}\;Var(b)>0$

We use the exact finite sample distribution of the likelihood ratio test statistic as derived by Crainiceanu & Ruppert (2004).

Value

A list of class htest containing the following components:

statistic the observed likelihood ratio
p p-value for the observed test statistic
method a character string indicating what type of test was performed and how many values were simulated to determine the critical value
sample the samples from the null distribution returned by LRTSim

Author(s)

Fabian Scheipl, updates for lme4.0-compatibility by Ben Bolker

References

Crainiceanu, C. and Ruppert, D. (2004) Likelihood ratio tests in linear mixed models with one variance component, Journal of the Royal Statistical Society: Series B,66,165–185.

Examples


library(nlme);
data(Orthodont);

##test for Sex:Age interaction and Subject-Intercept
mA<-lme(distance ~ Sex * I(age - 11), random = ~ 1| Subject,
  data = Orthodont, method = "ML")
m0<-lm(distance ~ Sex + I(age - 11), data = Orthodont)
summary(mA)
summary(m0)
exactLRT(m = mA, m0 = m0)

library(nlme);
data(Orthodont);

##test for Sex:Age interaction and Subject-Intercept
mA<-lme(distance ~ Sex * I(age - 11), random = ~ 1| Subject,
  data = Orthodont, method = "ML")
m0<-lm(distance ~ Sex + I(age - 11), data = Orthodont)
summary(mA)
summary(m0)
exactLRT(m = mA, m0 = m0)

Restricted Likelihood Ratio Tests for additive and linear mixed models

Description

This function provides an (exact) restricted likelihood ratio test based on simulated values from the finite sample distribution for testing whether the variance of a random effect is 0 in a linear mixed model with known correlation structure of the tested random effect and i.i.d. errors.

Usage

exactRLRT(
  m,
  mA = NULL,
  m0 = NULL,
  seed = NA,
  nsim = 10000,
  log.grid.hi = 8,
  log.grid.lo = -10,
  gridlength = 200,
  parallel = c("no", "multicore", "snow"),
  ncpus = 1L,
  cl = NULL
)
exactRLRT(
  m,
  mA = NULL,
  m0 = NULL,
  seed = NA,
  nsim = 10000,
  log.grid.hi = 8,
  log.grid.lo = -10,
  gridlength = 200,
  parallel = c("no", "multicore", "snow"),
  ncpus = 1L,
  cl = NULL
)

Arguments

`m`	The fitted model under the alternative or, for testing in models with multiple variance components, the reduced model containing only the random effect to be tested (see Details), an `lme`, `lmerMod` or `spm` object
`mA`	The full model under the alternative for testing in models with multiple variance components
`m0`	The model under the null for testing in models with multiple variance components
`seed`	input for `set.seed`
`nsim`	Number of values to simulate
`log.grid.hi`	Lower value of the grid on the log scale. See `exactRLRT`.
`log.grid.lo`	Lower value of the grid on the log scale. See `exactRLRT`.
`gridlength`	Length of the grid. See `exactLRT`.
`parallel`	The type of parallel operation to be used (if any). If missing, the default is "no parallelization").
`ncpus`	integer: number of processes to be used in parallel operation: typically one would chose this to the number of available CPUs. Defaults to 1, i.e., no parallelization.
`cl`	An optional parallel or snow cluster for use if parallel = "snow". If not supplied, a cluster on the local machine is created for the duration of the call.

Details

Testing in models with only a single variance component require only the first argument m. For testing in models with multiple variance components, the fitted model m must contain only the random effect set to zero under the null hypothesis, while mA and m0 are the models under the alternative and the null, respectively. For models with a single variance component, the simulated distribution is exact if the number of parameters (fixed and random) is smaller than the number of observations. Extensive simulation studies (see second reference below) confirm that the application of the test to models with multiple variance components is safe and the simulated distribution is correct as long as the number of parameters (fixed and random) is smaller than the number of observations and the nuisance variance components are not superfluous or very small. We use the finite sample distribution of the restricted likelihood ratio test statistic as derived by Crainiceanu & Ruppert (2004).

No simulation is performed if the observed test statistic is 0. (i.e., if the fit of the model fitted under the alternative is indistinguishable from the model fit under H0), since the p-value is always 1 in this case.

Value

A list of class htest containing the following components:

statistic the observed likelihood ratio
p p-value for the observed test statistic
method a character string indicating what type of test was performed and how many values were simulated to determine the critical value
sample the samples from the null distribution returned by RLRTSim

Author(s)

Fabian Scheipl, bug fixes by Andrzej Galecki, updates for lme4-compatibility by Ben Bolker

References

Crainiceanu, C. and Ruppert, D. (2004) Likelihood ratio tests in linear mixed models with one variance component, Journal of the Royal Statistical Society: Series B,66,165–185.

Greven, S., Crainiceanu, C., Kuechenhoff, H., and Peters, A. (2008) Restricted Likelihood Ratio Testing for Zero Variance Components in Linear Mixed Models, Journal of Computational and Graphical Statistics, 17 (4): 870–891.

Scheipl, F., Greven, S. and Kuechenhoff, H. (2008) Size and power of tests for a zero random effect variance or polynomial regression in additive and linear mixed models. Computational Statistics & Data Analysis, 52(7):3283–3299.

Examples


data(sleepstudy, package = "lme4")
mA <- lme4::lmer(Reaction ~ I(Days-4.5) + (1|Subject) + (0 + I(Days-4.5)|Subject), 
  data = sleepstudy)
m0 <- update(mA, . ~ . - (0 + I(Days-4.5)|Subject))
m.slope  <- update(mA, . ~ . - (1|Subject))
#test for subject specific slopes:
exactRLRT(m.slope, mA, m0)

library(mgcv)
data(trees)
#test quadratic trend vs. smooth alternative
m.q<-gamm(I(log(Volume)) ~ Height + s(Girth, m = 3), data = trees, 
  method = "REML")$lme
exactRLRT(m.q)
#test linear trend vs. smooth alternative
m.l<-gamm(I(log(Volume)) ~ Height + s(Girth, m = 2), data = trees, 
  method = "REML")$lme
exactRLRT(m.l)

data(sleepstudy, package = "lme4")
mA <- lme4::lmer(Reaction ~ I(Days-4.5) + (1|Subject) + (0 + I(Days-4.5)|Subject), 
  data = sleepstudy)
m0 <- update(mA, . ~ . - (0 + I(Days-4.5)|Subject))
m.slope  <- update(mA, . ~ . - (1|Subject))
#test for subject specific slopes:
exactRLRT(m.slope, mA, m0)

library(mgcv)
data(trees)
#test quadratic trend vs. smooth alternative
m.q<-gamm(I(log(Volume)) ~ Height + s(Girth, m = 3), data = trees, 
  method = "REML")$lme
exactRLRT(m.q)
#test linear trend vs. smooth alternative
m.l<-gamm(I(log(Volume)) ~ Height + s(Girth, m = 2), data = trees, 
  method = "REML")$lme
exactRLRT(m.l)

Extract the Design of a linear mixed model

Description

These functions extract various elements of the design of a fitted lme-, mer or lmerMod-Object. They are called by exactRLRT and exactLRT.

Usage

extract.lmeDesign(m)
extract.lmeDesign(m)

Arguments

`m`	a fitted `lme`- or `merMod`-Object

Value

a a list with components

Vr estimated covariance of the random effects divided by the estimated variance of the residuals
X design of the fixed effects
Z design of the random effects
sigmasq variance of the residuals
lambda ratios of the variances of the random effects and the variance of the residuals
y response variable

Author(s)

Fabian Scheipl, extract.lmerModDesign by Ben Bolker. Many thanks to Andrzej Galecki and Tomasz Burzykowski for bug fixes.

Examples


library(nlme)
design <- extract.lmeDesign(lme(distance ~ age + Sex, data = Orthodont, 
                             random = ~ 1))
str(design)

library(nlme)
design <- extract.lmeDesign(lme(distance ~ age + Sex, data = Orthodont, 
                             random = ~ 1))
str(design)

Simulation of the (Restricted) Likelihood Ratio Statistic

Description

These functions simulate values from the (exact) finite sample distribution of the (restricted) likelihood ratio statistic for testing the presence of the variance component (and restrictions of the fixed effects) in a simple linear mixed model with known correlation structure of the random effect and i.i.d. errors. They are usually called by exactLRT or exactRLRT.

Usage

LRTSim(
  X,
  Z,
  q,
  sqrt.Sigma,
  seed = NA,
  nsim = 10000,
  log.grid.hi = 8,
  log.grid.lo = -10,
  gridlength = 200,
  parallel = c("no", "multicore", "snow"),
  ncpus = 1L,
  cl = NULL
)
LRTSim(
  X,
  Z,
  q,
  sqrt.Sigma,
  seed = NA,
  nsim = 10000,
  log.grid.hi = 8,
  log.grid.lo = -10,
  gridlength = 200,
  parallel = c("no", "multicore", "snow"),
  ncpus = 1L,
  cl = NULL
)

Arguments

`X`	The fixed effects design matrix of the model under the alternative
`Z`	The random effects design matrix of the model under the alternative
`q`	The number of parameters restrictions on the fixed effects (see Details)
`sqrt.Sigma`	The upper triangular Cholesky factor of the correlation matrix of the random effect
`seed`	Specify a seed for `set.seed`
`nsim`	Number of values to simulate
`log.grid.hi`	Lower value of the grid on the log scale. See Details
`log.grid.lo`	Lower value of the grid on the log scale. See Details
`gridlength`	Length of the grid for the grid search over lambda. See Details
`parallel`	The type of parallel operation to be used (if any). If missing, the default is "no parallelization").
`ncpus`	integer: number of processes to be used in parallel operation: typically one would chose this to the number of available CPUs. Defaults to 1, i.e., no parallelization.
`cl`	An optional parallel or snow cluster for use if parallel = "snow". If not supplied, a cluster on the local machine is created for the duration of the call.

Details

The model under the alternative must be a linear mixed model $y=X\beta+Zb+\varepsilon$ with a single random effect $b$ with known correlation structure $Sigma$ and i.i.d errors. The simulated distribution of the likelihood ratio statistic was derived by Crainiceanu & Ruppert (2004). The simulation algorithm uses a grid search over a log-regular grid of values of $\lambda=\frac{Var(b)}{Var(\varepsilon)}$ to maximize the likelihood under the alternative for nsim realizations of $y$ drawn under the null hypothesis. log.grid.hi and log.grid.lo are the lower and upper limits of this grid on the log scale. gridlength is the number of points on the grid.\ These are just wrapper functions for the underlying C code.

Value

A vector containing the the simulated values of the (R)LRT under the null, with attribute 'lambda' giving $\arg\min(f(\lambda))$ (see Crainiceanu, Ruppert (2004)) for the simulations.

Author(s)

Fabian Scheipl; parallelization code adapted from boot package

References

Crainiceanu, C. and Ruppert, D. (2004) Likelihood ratio tests in linear mixed models with one variance component, Journal of the Royal Statistical Society: Series B,66,165–185.

Scheipl, F. (2007) Testing for nonparametric terms and random effects in structured additive regression. Diploma thesis (unpublished).

Scheipl, F., Greven, S. and Kuechenhoff, H (2008) Size and power of tests for a zero random effect variance or polynomial regression in additive and linear mixed models, Computational Statistics & Data Analysis, 52(7):3283-3299

Examples


library(lme4)
g <- rep(1:10, e = 10)
x <- rnorm(100)
y <- 0.1 * x + rnorm(100)
m <- lmer(y ~ x + (1|g), REML=FALSE)
m0 <- lm(y ~ 1)

(obs.LRT <- 2*(logLik(m)-logLik(m0)))
X <- getME(m,"X")
Z <- t(as.matrix(getME(m,"Zt")))
sim.LRT <- LRTSim(X, Z, 1, diag(10))
(pval <- mean(sim.LRT > obs.LRT))

library(lme4)
g <- rep(1:10, e = 10)
x <- rnorm(100)
y <- 0.1 * x + rnorm(100)
m <- lmer(y ~ x + (1|g), REML=FALSE)
m0 <- lm(y ~ 1)

(obs.LRT <- 2*(logLik(m)-logLik(m0)))
X <- getME(m,"X")
Z <- t(as.matrix(getME(m,"Zt")))
sim.LRT <- LRTSim(X, Z, 1, diag(10))
(pval <- mean(sim.LRT > obs.LRT))

Package 'RLRsim'

Help Index

R package for fast and exact (restricted) likelihood ratio tests for mixed and additive models.

Description

Author(s)

Likelihood Ratio Tests for simple linear mixed models

Description

Usage

Arguments

Details

Value

Author(s)

References

See Also

Examples

Restricted Likelihood Ratio Tests for additive and linear mixed models

Description

Usage

Arguments

Details

Value

Author(s)

References

See Also

Examples

Extract the Design of a linear mixed model

Description

Usage

Arguments

Value

Author(s)

Examples

Simulation of the (Restricted) Likelihood Ratio Statistic

Description

Usage

Arguments

Details

Value

Author(s)

References

See Also

Examples