tests/testthat/test-parametric-nlme.R

In lmeresampler: Bootstrap Methods for Nested Linear Mixed-Effects Models

library(nlme, quietly = TRUE)

data(Socatt, package = "mlmRev")
Socatt$religion <- relevel(Socatt$religion, ref = "none")
Socatt$rv <- as.numeric(as.character(Socatt$numpos))
Socatt$rv <- scale(Socatt$rv) # a plot shows this is clearly non-normal



# ==============================================================================
context("parametric bootstrap (lme)")
# ==============================================================================

mySumm <- function(.) { 
  c(beta = fixef(.), sigma = as.numeric(.$sigma), sig01 = as.numeric(VarCorr(.)[1,2]))
}

nsim <- 10

test_that("two-level additive random intercept model",{
  skip_on_cran()
  ## See p. 31 of Goldstein's book
  vcmodA <- lme(mathAge11 ~ mathAge8 + gender + class,
                random = ~ 1 | school, data = jsp728)
  
  orig.stats <- mySumm(vcmodA)
  
  boo <- parametric_bootstrap(model = vcmodA, .f = mySumm, B = nsim)
  
  expect_equal(class(boo), "lmeresamp")
  expect_equal(boo$observed, orig.stats)   
  expect_equal(unname(boo$stats$observed), unname(orig.stats))
  expect_equal(nrow(boo$replicates), nsim)
  expect_equal(ncol(boo$replicates), length(orig.stats))
  expect_equal(boo$B, nsim)
  expect_equal(boo$type, "parametric")
  expect_equal(boo$.f, mySumm)
})

# ------------------------------------------------------------------------------

test_that("two-level random intercept model without interaction",{
  skip_on_cran()
  ## See p. 97 of Goldstein's book
  rimod <- lme(mathAge11 ~ mathAge8c + gender + class,
               random = ~ 1 | school, data = jsp728)
  
  orig.stats <- mySumm(rimod)
  boo <- parametric_bootstrap(model = rimod, .f = mySumm, B = nsim)
  
  expect_equal(class(boo), "lmeresamp")
  expect_equal(boo$observed, orig.stats)   
  expect_equal(unname(boo$stats$observed), unname(orig.stats))
  expect_equal(nrow(boo$replicates), nsim)
  expect_equal(ncol(boo$replicates), length(orig.stats))
  expect_equal(boo$B, nsim)
  expect_equal(boo$type, "parametric")
  expect_equal(boo$.f, mySumm)
})


test_that("two-level random intercept model with interaction",{
  skip_on_cran()
  ## See p. 34 of Goldstein's book
  vcmodC <- lme(mathAge11 ~ mathAge8 * schoolMathAge8 + gender + class, 
                random = ~ 1 | school, data = jsp728)
  
  orig.stats <- mySumm(vcmodC)
  boo <- parametric_bootstrap(model = vcmodC, .f = mySumm, B = nsim)
  
  expect_equal(class(boo), "lmeresamp")
  expect_equal(boo$observed, orig.stats)   
  expect_equal(unname(boo$stats$observed), unname(orig.stats))
  expect_equal(nrow(boo$replicates), nsim)
  expect_equal(ncol(boo$replicates), length(orig.stats))
  expect_equal(boo$B, nsim)
  expect_equal(boo$type, "parametric")
  expect_equal(boo$.f, mySumm)
})

# ------------------------------------------------------------------------------

test_that("two-level random coefficient model with interaction",{
  skip_on_cran()
  ## See p. 35 of Goldstein's book
  rcmod <- lme(mathAge11 ~ mathAge8c * schoolMathAge8 + gender + class,
               random = ~ mathAge8c | school, data = jsp728)
  
  orig.stats <- mySumm(rcmod)
  boo <- parametric_bootstrap(model = rcmod, .f = mySumm, B = nsim)
  
  expect_equal(class(boo), "lmeresamp")
  expect_equal(boo$observed, orig.stats)   
  expect_equal(unname(boo$stats$observed), unname(orig.stats))
  expect_equal(nrow(boo$replicates), nsim)
  expect_equal(ncol(boo$replicates), length(orig.stats))
  expect_equal(boo$B, nsim)
  expect_equal(boo$type, "parametric")
  expect_equal(boo$.f, mySumm)
})

# ------------------------------------------------------------------------------

# test_that("three-level random intercept model",{
#   rmA <- lme(rv ~ religion + year, random = ~ 1 | district/respond, data = Socatt)
#   
#   
#   mySumm <- function(.) { 
#     c(beta = fixef(.), sigma = as.numeric(.$sigma), sig01 = as.numeric(VarCorr(.)[1,2]))
#   }
#   
#   orig.stats <- mySumm(rmA)
#   boo <- parametric_bootstrap(model = rmA, .f = mySumm, B = nsim)
#   
#   expect_equal(class(boo), "boot")
#   expect_equal(boo$t0, orig.stats)
#   expect_equal(nrow(boo$t), nsim)
#   expect_equal(ncol(boo$t), length(orig.stats))
#   expect_equal(boo$B, nsim)
#   expect_equal(boo$sim, "parametric")
#   expect_equal(boo$statistic, mySumm)
# })