[R-pkg-team] Bug#921938: lme4 breaks r-cran-lmertest autopkgtest
Paul Gevers
elbrus at debian.org
Sun Feb 10 12:35:53 GMT 2019
Source: lme4, r-cran-lmertest
Control: found -1 lme4/1.1-20-1
Control: found -1 r-cran-lmertest/3.0-1-2
Severity: important
X-Debbugs-CC: debian-ci at lists.debian.org
User: debian-ci at lists.debian.org
Usertags: breaks needs-update
Dear maintainers,
With a recent upload of lme4 the autopkgtest of r-cran-lmertest fails in
testing when that autopkgtest is run with the binary packages of lme4
from unstable. It passes when run with only packages from testing. In
tabular form:
pass fail
lme4 from testing 1.1-20-1
r-cran-lmertest from testing 3.0-1-2
all others from testing from testing
I copied some of the output at the bottom of this report.
Currently this regression is blocking the migration of lme4 to testing
[1]. Due to the nature of this issue, I filed this bug report against
both packages. Can you please investigate the situation and reassign the
bug to the right package? If needed, please change the bug's severity.
More information about this bug and the reason for filing it can be found on
https://wiki.debian.org/ContinuousIntegration/RegressionEmailInformation
Paul
[1] https://qa.debian.org/excuses.php?package=lme4
https://ci.debian.net/data/autopkgtest/testing/amd64/r/r-cran-lmertest/1896879/log.gz
autopkgtest [04:42:34]: test run-unit-test: [-----------------------
BEGIN TEST test_a_utils.R
R version 3.5.2 (2018-12-20) -- "Eggshell Igloo"
Copyright (C) 2018 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)
R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.
R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.
> # test_a_utils.R
>
> library(lmerTest)
Loading required package: lme4
Loading required package: Matrix
Attaching package: 'lmerTest'
The following object is masked from 'package:lme4':
lmer
The following object is masked from 'package:stats':
step
>
> # test safeDeparse() - equivalence and differences to deparse():
> deparse_args <- formals(deparse)
> safeDeparse_args <- formals(lmerTest:::safeDeparse)
> stopifnot(
+ all.equal(names(deparse_args), names(safeDeparse_args)),
+ all.equal(deparse_args[!names(deparse_args) %in% c("control",
"width.cutoff")],
+ safeDeparse_args[!names(safeDeparse_args) %in%
c("control", "width.cutoff")]),
+ all.equal(deparse_args[["width.cutoff"]], 60L),
+ all(eval(safeDeparse_args[["control"]]) %in%
eval(deparse_args[["control"]])),
+ all.equal(safeDeparse_args[["width.cutoff"]], 500L)
+ )
>
>
BEGIN TEST test_anova.R
R version 3.5.2 (2018-12-20) -- "Eggshell Igloo"
Copyright (C) 2018 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)
R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.
R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.
> # test_anova.R
> library(lmerTest)
Loading required package: lme4
Loading required package: Matrix
Attaching package: 'lmerTest'
The following object is masked from 'package:lme4':
lmer
The following object is masked from 'package:stats':
step
>
> # WRE says "using if(requireNamespace("pkgname")) is preferred, if
possible."
> # even in tests:
> assertError <- function(expr, ...)
+ if(requireNamespace("tools")) tools::assertError(expr, ...) else
invisible()
> assertWarning <- function(expr, ...)
+ if(requireNamespace("tools")) tools::assertWarning(expr, ...) else
invisible()
>
> # Kenward-Roger only available with pbkrtest and only then validated
in R >= 3.3.3
> # (faulty results for R < 3.3.3 may be due to unstated dependencies in
pbkrtest)
> has_pbkrtest <- requireNamespace("pbkrtest", quietly = TRUE) &&
getRversion() >= "3.3.3"
>
> data("sleepstudy", package="lme4")
> TOL <- 1e-4
>
> ####################################
> ## Basic anova tests
> ####################################
>
> m <- lmer(Reaction ~ Days + (Days | Subject), sleepstudy)
>
> ####### ddf argument:
> (an1 <- anova(m)) # Also testing print method.
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, ddf="Satterthwaite"))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2b <- anova(m, ddf="Satterthwaite", type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2c <- anova(m, ddf="Satterthwaite", type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an2, tolerance=TOL)
+ ))
> (an3 <- anova(m, ddf="Sat")) ## Abbreviated argument
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an3, tolerance=TOL)
+ ))
> if(has_pbkrtest) {
+ (anova(m, ddf="Kenward-Roger"))
+ (anova(m, ddf="Kenward-Roger", type=3))
+ }
Type III Analysis of Variance Table with Kenward-Roger's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 17 45.843 3.268e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an1 <- anova(m, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
Days 1 30024 30024 45.843
> (an2 <- anova(m, ddf="lme4", type=3)) # 'type' is ignored with ddf="lme4"
Analysis of Variance Table
Df Sum Sq Mean Sq F value
Days 1 30024 30024 45.843
> stopifnot(isTRUE(
+ all.equal(an1, an2, tolerance=TOL)
+ ))
> res <- assertError(anova(m, ddf="KR")) ## Error on incorrect arg.
> stopifnot(
+ grepl("'arg' should be one of ", unlist(res[[1]])$message)
+ )
>
> ## lme4 method:
> an1 <- anova(m, ddf="lme4")
> an2 <- anova(as(m, "lmerMod"))
> stopifnot(isTRUE(
+ all.equal(an1, an2, tolerance=TOL)
+ ))
>
> ###### type argument:
> (an1 <- anova(m, type="1")) # valid type arg.
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type="I")) # same
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an2, tolerance=TOL)
+ ))
> (an3 <- anova(m, type=1)) # Not strictly valid, but accepted
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an3, tolerance=TOL)
+ ))
>
> (an1 <- anova(m, type="2")) # valid type arg.
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type="II")) # same
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an2, tolerance=TOL)
+ ))
> (an3 <- anova(m, type=3)) # Not strictly valid, but accepted
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an3, check.attributes=FALSE, tolerance=TOL)
+ ))
>
> (an1 <- anova(m, type="3")) # valid type arg.
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type="III")) # same
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an2, tolerance=TOL)
+ ))
> (an3 <- anova(m, type=3)) # Not strictly valid, but accepted
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 30024 30024 1 16.995 45.843 3.273e-06 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(isTRUE(
+ all.equal(an1, an3, tolerance=TOL)
+ ))
> assertError(anova(m, type=0)) # Not valid arg.
> assertError(anova(m, type="i")) # Not valid arg.
>
> ####### Model comparison:
> fm <- lm(Reaction ~ Days, sleepstudy)
> (an <- anova(m, fm))
refitting model(s) with ML (instead of REML)
Data: sleepstudy
Models:
fm: Reaction ~ Days
m: Reaction ~ Days + (Days | Subject)
Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq)
fm 3 1906.3 1915.9 -950.15 1900.3
m 6 1763.9 1783.1 -875.97 1751.9 148.35 3 < 2.2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(
+ nrow(an) == 2L,
+ rownames(an)[2] == "m"
+ )
>
> m2 <- lmer(Reaction ~ Days + I(Days^2) + (Days | Subject), sleepstudy)
> (an <- anova(m, m2, refit=FALSE))
Data: sleepstudy
Models:
m: Reaction ~ Days + (Days | Subject)
m2: Reaction ~ Days + I(Days^2) + (Days | Subject)
Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq)
m 6 1755.6 1774.8 -871.81 1743.6
m2 7 1756.8 1779.2 -871.41 1742.8 0.8127 1 0.3673
> stopifnot(
+ nrow(an) == 2L,
+ rownames(an)[1] == "m"
+ )
>
>
> ####################################
> ## Example with factor fixef:
> ####################################
>
> ## 'temp' is continuous, 'temperature' an ordered factor with 6 levels
> data("cake", package="lme4")
> m <- lmer(angle ~ recipe * temp + (1|recipe:replicate), cake)
> (an <- anova(m))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 4.00 2.00 2 254.02 0.0957 0.9088
temp 1966.71 1966.71 1 222.00 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222.00 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an_lme4 <- anova(m, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
recipe 2 10.39 5.20 0.2488
temp 1 1966.71 1966.71 94.1632
recipe:temp 2 1.74 0.87 0.0417
>
> if(has_pbkrtest) {
+ (an_KR <- anova(m, ddf="Kenward-Roger"))
+ # res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ # an_lme4[, c("Sum Sq", "Mean Sq", "F value")])
+ # stopifnot(isTRUE(res))
+ res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ an_KR[, c("Sum Sq", "Mean Sq", "F value")],
tolerance=TOL)
+ stopifnot(isTRUE(res))
+ }
> stopifnot(all.equal(c(2, 1, 2), an$NumDF, tol=1e-6),
+ all.equal(c(254.0157612, 222, 222), an$DenDF, tol=TOL))
>
> an3 <- anova(m, type=3)
> an2 <- anova(m, type=2)
> an1 <- anova(m, type=1)
>
> ## Data is balanced, so Type II and III should be identical:
> ## One variable is continuous, so Type I and II/III are different:
> stopifnot(
+ isTRUE(all.equal(an3, an2, check.attributes=FALSE, tolerance=TOL)),
+ !isTRUE(all.equal(an1, an2, check.attributes=FALSE, tolerance=1e-8))
+ )
>
> # Using an ordered factor:
> m <- lmer(angle ~ recipe * temperature + (1|recipe:replicate), cake)
> (an1 <- anova(m, type=1))
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.19 5.09 2 42 0.2488 0.7809
temperature 2100.30 420.06 5 210 20.5199 <2e-16 ***
recipe:temperature 205.98 20.60 10 210 1.0062 0.4393
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.19 5.09 2 42 0.2488 0.7809
temperature 2100.30 420.06 5 210 20.5199 <2e-16 ***
recipe:temperature 205.98 20.60 10 210 1.0062 0.4393
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> # Type 3 is also available with ordered factors:
> (an3 <- anova(m, type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.19 5.09 2 42 0.2488 0.7809
temperature 2100.30 420.06 5 210 20.5199 <2e-16 ***
recipe:temperature 205.98 20.60 10 210 1.0062 0.4393
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> ## Balanced data and only factors: Type I, II and III should be the same:
> stopifnot(
+ isTRUE(all.equal(an1, an2, check.attributes=FALSE, tolerance=TOL)),
+ isTRUE(all.equal(an1, an3, check.attributes=FALSE, tolerance=TOL))
+ )
>
> (an <- anova(m, type=1))
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.19 5.09 2 42 0.2488 0.7809
temperature 2100.30 420.06 5 210 20.5199 <2e-16 ***
recipe:temperature 205.98 20.60 10 210 1.0062 0.4393
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an_lme4 <- anova(m, type=1, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
recipe 2 10.19 5.09 0.2488
temperature 5 2100.30 420.06 20.5199
recipe:temperature 10 205.98 20.60 1.0062
> res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ an_lme4[, c("Sum Sq", "Mean Sq", "F value")],
tolerance=TOL)
> stopifnot(isTRUE(res))
> if(has_pbkrtest) {
+ (an_KR <- anova(m, type=1, ddf="Kenward-Roger"))
+ res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ an_KR[, c("Sum Sq", "Mean Sq", "F value")],
tolerance=TOL)
+ stopifnot(isTRUE(res))
+ }
> stopifnot(all.equal(c(2, 5, 10), an$NumDF, tolerance=TOL),
+ all.equal(c(42, 210, 210), an$DenDF, tolerance=TOL))
>
> ########
> ## Make case with balanced unordered factors:
> cake2 <- cake
> cake2$temperature <- factor(cake2$temperature, ordered = FALSE)
> # str(cake2)
> stopifnot(
+ !is.ordered(cake2$temperature)
+ )
> m <- lmer(angle ~ recipe * temperature + (1|recipe:replicate), cake2)
> (an1 <- anova(m, type=1))
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.19 5.09 2 42 0.2488 0.7809
temperature 2100.30 420.06 5 210 20.5199 <2e-16 ***
recipe:temperature 205.98 20.60 10 210 1.0062 0.4393
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.19 5.09 2 42 0.2488 0.7809
temperature 2100.30 420.06 5 210 20.5199 <2e-16 ***
recipe:temperature 205.98 20.60 10 210 1.0062 0.4393
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an3 <- anova(m, type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.19 5.09 2 42 0.2488 0.7809
temperature 2100.30 420.06 5 210 20.5199 <2e-16 ***
recipe:temperature 205.98 20.60 10 210 1.0062 0.4393
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> ## Balanced data and only factors: Type I, II, and III should be the same:
> stopifnot(
+ isTRUE(all.equal(an1, an2, check.attributes=FALSE, tolerance=TOL)),
+ isTRUE(all.equal(an3, an2, check.attributes=FALSE, tolerance=TOL))
+ )
> ########
>
> # No intercept:
> m <- lmer(angle ~ 0 + recipe * temp + (1|recipe:replicate), cake)
> (an <- anova(m, type=1))
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 21442.9 7147.6 3 42 342.2200 <2e-16 ***
temp 1966.7 1966.7 1 222 94.1632 <2e-16 ***
recipe:temp 1.7 0.9 2 222 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 4.48 1.49 3 254.02 0.0714 0.9752
temp 1966.71 1966.71 1 222.00 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222.00 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 4.48 1.49 3 254.02 0.0714 0.9752
temp 1966.71 1966.71 1 222.00 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222.00 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> if(has_pbkrtest)
+ (an_KR <- anova(m, ddf="Kenward-Roger"))
Type III Analysis of Variance Table with Kenward-Roger's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 4.48 1.49 3 254.02 0.0714 0.9752
temp 1966.71 1966.71 1 222.00 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222.00 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an_lme4 <- anova(m, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
recipe 3 21442.9 7147.6 342.2200
temp 1 1966.7 1966.7 94.1632
recipe:temp 2 1.7 0.9 0.0417
> res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ an_lme4[, c("Sum Sq", "Mean Sq", "F value")],
tolerance=TOL)
> stopifnot(isTRUE(res))
>
> # ML-fit:
> m <- lmer(angle ~ recipe * temp + (1|recipe:replicate), cake, REML=FALSE)
> (an <- anova(m, type=1))
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.99 5.49 2 45 0.2666 0.7672
temp 1966.71 1966.71 1 225 95.4357 <2e-16 ***
recipe:temp 1.74 0.87 2 225 0.0423 0.9586
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> if(has_pbkrtest)
+ assertError(an <- anova(m, ddf="Kenward-Roger")) # KR fits should be
REML
> (an_lme4 <- anova(m, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
recipe 2 10.99 5.49 0.2666
temp 1 1966.71 1966.71 95.4357
recipe:temp 2 1.74 0.87 0.0423
> res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ an_lme4[, c("Sum Sq", "Mean Sq", "F value")],
tolerance=TOL)
> stopifnot(isTRUE(res))
>
> ####################################
> ## Using contr.sum:
> ####################################
>
> m <- lmer(angle ~ recipe * temp + (1|recipe:replicate), cake,
+ contrasts = list('recipe' = "contr.sum"))
> (an <- anova(m, type=1))
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.39 5.20 2 42 0.2488 0.7809
temp 1966.71 1966.71 1 222 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an2 <- anova(m, type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 4.00 2.00 2 254.02 0.0957 0.9088
temp 1966.71 1966.71 1 222.00 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222.00 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an3 <- anova(m, type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 4.00 2.00 2 254.02 0.0957 0.9088
temp 1966.71 1966.71 1 222.00 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222.00 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> stopifnot(
+ isTRUE(all.equal(an2, an3, check.attributes=FALSE, tolerance=TOL))
+ )
> if(has_pbkrtest)
+ (an_KR <- anova(m, type=1, ddf="Kenward-Roger"))
Type I Analysis of Variance Table with Kenward-Roger's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
recipe 10.39 5.20 2 42 0.2488 0.7809
temp 1966.71 1966.71 1 222 94.1632 <2e-16 ***
recipe:temp 1.74 0.87 2 222 0.0417 0.9592
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an_lme4 <- anova(m, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
recipe 2 10.39 5.20 0.2488
temp 1 1966.71 1966.71 94.1632
recipe:temp 2 1.74 0.87 0.0417
> res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ an_lme4[, c("Sum Sq", "Mean Sq", "F value")],
tolerance=TOL)
> stopifnot(isTRUE(res))
>
>
> ####################################
> ## Example with continuous fixef:
> ####################################
>
> # Example with no fixef:
> m <- lmer(Reaction ~ -1 + (Days | Subject), sleepstudy)
> # m <- lmer(Reaction ~ 0 + (Days | Subject), sleepstudy) # alternative
> stopifnot(length(fixef(m)) == 0L)
> (an <- anova(m, type=1))
Type I Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
> (an_2 <- anova(m, type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
> (an_3 <- anova(m, type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
> stopifnot(nrow(an) == 0L,
+ nrow(an_2) == 0L,
+ nrow(an_3) == 0L)
> # anova(m, ddf="lme4") # Bug in lme4 it seems
> if(has_pbkrtest) {
+ (an_KR <- anova(m, ddf="Kenward-Roger"))
+ stopifnot(
+ nrow(an_KR) == 0L
+ )
+ }
>
> # Example with intercept only:
> m <- lmer(Reaction ~ (Days | Subject), sleepstudy)
> # m <- lmer(Reaction ~ 1 + (Days | Subject), sleepstudy) # alternative
> stopifnot(length(fixef(m)) == 1L,
+ names(fixef(m)) == "(Intercept)")
> (an <- anova(m))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
> (an_2 <- anova(m, type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
> (an_3 <- anova(m, type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
> (an_lme4 <- anova(m, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
> stopifnot(nrow(an) == 0L,
+ nrow(an_2) == 0L,
+ nrow(an_3) == 0L,
+ nrow(an_lme4) == 0L)
> if(has_pbkrtest) {
+ (an_KR <- anova(m, ddf="Kenward-Roger"))
+ stopifnot(
+ nrow(an_KR) == 0L
+ )
+ }
>
> # Example with 1 fixef without intercept:
> m <- lmer(Reaction ~ Days - 1 + (Days | Subject), sleepstudy)
Warning message:
In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge with max|grad| = 0.00416642 (tol = 0.002,
component 1)
> # m <- lmer(Reaction ~ 0 + Days + (Days | Subject), sleepstudy) #
alternative
> stopifnot(length(fixef(m)) == 1L,
+ names(fixef(m)) == "Days")
> (an <- anova(m))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 93779 93779 1 16.995 143.19 1.054e-09 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an_2 <- anova(m, type=2))
Type II Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 93779 93779 1 16.995 143.19 1.054e-09 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an_3 <- anova(m, type=3))
Type III Analysis of Variance Table with Satterthwaite's method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Days 93779 93779 1 16.995 143.19 1.054e-09 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> (an_lme4 <- anova(m, ddf="lme4"))
Analysis of Variance Table
Df Sum Sq Mean Sq F value
Days 1 93779 93779 143.19
> stopifnot(nrow(an) == 1L,
+ nrow(an_2) == 1L,
+ nrow(an_3) == 1L,
+ nrow(an_lme4) == 1L)
> if(has_pbkrtest) {
+ (an_KR <- anova(m, ddf="Kenward-Roger"))
+ stopifnot(
+ nrow(an_KR) == 1L
+ )
+ }
>
> res <- all.equal(an[, c("Sum Sq", "Mean Sq", "F value")],
+ an_lme4[, c("Sum Sq", "Mean Sq", "F value")],
tolerance=TOL)
> stopifnot(isTRUE(res))
> stopifnot(isTRUE(all.equal(
+ c(1, 17), unname(unlist(an[, c("NumDF", "DenDF")])), tolerance=TOL
+ )))
Error: isTRUE(all.equal(c(1, 17), unname(unlist(an[, c("NumDF",
"DenDF")])), .... is not TRUE
Execution halted
autopkgtest [04:42:39]: test run-unit-test: -----------------------]
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://alioth-lists.debian.net/pipermail/r-pkg-team/attachments/20190210/116b08a5/attachment-0001.sig>
More information about the R-pkg-team
mailing list