tests/testthat/test-generateLongData.R

In latrend: A Framework for Clustering Longitudinal Data

context('generateLongData')
rngReset()

test_that('generateLongData without fixed and random effects', {
  dt = generateLongData(
    sizes=c(1, 2),
    fixed=Value ~ 0,
    cluster=~1,
    random=~0,
    id='Id',
    data=data.frame(Time=c(0,1)),
    fixedCoefs=NULL,
    clusterCoefs=cbind(1,2),
    randomScales=NULL
  )

  expect_is(dt, 'data.frame')
  expect_equal(nrow(dt), 6)
  expect_true(all(dt$Mu.fixed == 0))
  expect_equal(dt$Mu.class, rep(1:2, c(2, 4)))
  expect_equal(dt$Mu, rep(1:2, c(2, 4)))
  expect_false(has_name(dt, 'Mu.random'))
})

test_that('generateLongData with fixed effects', {
  dt = generateLongData(
    sizes=c(1, 1),
    fixed=Value ~ Time,
    cluster=~1,
    random=~0,
    id='Id',
    data=data.frame(Time=c(0,.5,1)),
    fixedCoefs=c(1,2),
    clusterCoefs=cbind(1,2),
    randomScales=NULL
  )

  expect_is(dt, 'data.frame')
  expect_length(dt, 7)
  expect_equal(nrow(dt), 6)
  expect_equal(dt$Mu.fixed, rep(1:3, 2))
  expect_equal(dt$Mu.class, rep(1:2, each=3))
  expect_equal(dt$Mu, c(2:4, 3:5))
  expect_false(has_name(dt, 'Mu.random'))
})

test_that('generateLongData with random intercept', {
  set.seed(1)
  dt = generateLongData(
    sizes=c(10, 10),
    fixed=Value ~ Time,
    cluster=~1,
    random=~1,
    id='Id',
    data=data.frame(Time=c(0,.5,1)),
    fixedCoefs=c(1,2),
    clusterCoefs=cbind(1,2),
    randomScales=cbind(.1,10)
  )

  expect_true(has_name(dt, 'Mu.random'))
  expect_equal(dt[Time == 0, round(log10(sd(Mu))), by=Class]$V1, c(-1, 1))
})

test_that('generateLongData with scalar random scale', {
  set.seed(1)
  dt = generateLongData(
    sizes=c(10, 10),
    fixed=Value ~ Time,
    cluster=~1,
    random=~1,
    id='Id',
    data=data.frame(Time=c(0,.5,1)),
    fixedCoefs=c(1,2),
    clusterCoefs=cbind(1,2),
    randomScales=.1
  )

  expect_true(has_name(dt, 'Mu.random'))
  expect_equal(dt[Time == 0, round(log10(sd(Mu))), by=Class]$V1, c(-1, -1))
})

test_that('generateLongData with multiple terms', {
  set.seed(1)
  dt = generateLongData(
    sizes=c(1, 1),
    fixed=Value ~ poly(Time, 2, raw=TRUE),
    cluster=~poly(Time, 2, raw=TRUE),
    random=~poly(Time, 2, raw=TRUE),
    id='Id',
    data=data.frame(Time=c(0,.5,1)),
    fixedCoefs=c(1,2,3),
    clusterCoefs=cbind(c(1,1,1),c(2,2,2)),
    randomScales=cbind(c(.1,1,2), c(10,1,1))
  )

  expect_equal(nrow(dt), 6)
  expect_length(dt, 10)
  expect_equal(dt$Mu.class, c(1, 1.75, 3, 2, 3.5, 6))
})