tests/testthat/test-mhglm.R

context("mhglm")

expect_equal_tol1 <- function(...) expect_equal(..., tolerance = 1e-1)

test_that("succeeds on sleepstudy", {
    model <- mhglm(Reaction ~ Days + (Days | Subject), data = lme4::sleepstudy)

    # fixef
    fixef0 <- c("(Intercept)" = 251.4, "Days" = 10.5)
    expect_equal_tol1(fixef(model), fixef0)

    # vcov
    vcov0 <- matrix(c(44.0, -1.4, -1.4, 2.3), 2, 2)
    rownames(vcov0) <- colnames(vcov0) <- c("(Intercept)", "Days")
    expect_equal_tol1(vcov(model), vcov0)

    # VarCorr
    varcor0 <- matrix(c(565.5, 11.1, 11.1, 32.7), 2, 2)
    rownames(varcor0) <- colnames(varcor0) <- c("(Intercept)", "Days")
    varcor <- VarCorr(model)[["Subject"]]
    expect_equal(attr(varcor, "stddev"), sqrt(diag(varcor)))
    expect_equal(as.vector(attr(varcor, "correlation")),
                as.vector(t(varcor / attr(varcor, "stddev"))
                                 / attr(varcor, "stddev")))
    attr(varcor, "stddev") <- NULL
    attr(varcor, "correlation") <- NULL
    expect_equal_tol1(varcor, varcor0)

    # ranef
    ranef0 <- matrix(c( 2.8, -40.0, -38.4, 22.8, 21.5,  8.8,  16.4, -7.0,  -1.0,
                       34.7, -24.6, -12.3,  4.3, 20.6,  3.3, -24.7,  0.7,  12.1,
                        9.1,  -8.6,  -5.5, -4.7, -2.9, -0.2,  -0.2,  1.0, -10.6,
                        8.6,   1.1,   6.5, -3.0,  3.6,  0.9,   4.7, -1.0,   1.3
                       ), 18, 2)
    rownames(ranef0) <- as.character(c(308, 309, 310, 330, 331, 332, 333, 334,
                                       335, 337, 349, 350, 351, 352, 369, 370,
                                       371, 372))
    colnames(ranef0) <- c("(Intercept)", "Days")
    expect_equal_tol1(as.matrix(ranef(model)[["Subject"]]), ranef0)
})

test_that("succeeds on sleepstudy in parallel", {
    model <- mhglm(Reaction ~ Days + (Days | Subject), data = lme4::sleepstudy)
    model_p <- mhglm(Reaction ~ Days + (Days | Subject), data = lme4::sleepstudy,
                     control = list(parallel = TRUE))

    expect_equal(fixef(model_p), fixef(model))
    expect_equal(vcov(model_p), vcov(model))
    expect_equal(VarCorr(model_p), VarCorr(model))
    expect_equal(ranef(model_p), ranef(model))
})


test_that("succeeds on cbpp", {
    suppressWarnings({
        model <- mhglm(cbind(incidence, size - incidence) ~ period + (period | herd),
                       data=lme4::cbpp, family=binomial)
    })

    # fixef
    fixef0 <- c("(Intercept)" = -1.2, "period2" = -0.8, "period3" = -0.9,
                "period4" = -1.1)
    expect_equal_tol1(fixef(model), fixef0)

    # vcov
    vcov0 <- matrix(c( 0.1, -0.1, -0.1, -0.1, -0.1, 0.2, 0.1, 0.0,
                      -0.1,  0.1,  0.2,  0.0, -0.1, 0.0, 0.0, 0.1), 4, 4)
    rownames(vcov0) <- colnames(vcov0) <- c("(Intercept)", "period2",
                                            "period3", "period4")
    expect_equal_tol1(vcov(model), vcov0)

    # VarCorr
    varcor0 <- matrix(c( 0.7, -0.7, -0.4, -0.5, -0.7, 1.0, 1.0, 0.3,
                        -0.4,  1.0,  1.4,  0.0, -0.5, 0.3, 0.0, 0.3), 4, 4)
    rownames(varcor0) <- colnames(varcor0) <- c("(Intercept)", "period2",
                                                "period3", "period4")
    varcor <- VarCorr(model)[["herd"]]
    expect_equal(attr(varcor, "stddev"), sqrt(diag(varcor)))
    expect_equal(as.vector(attr(varcor, "correlation")),
                as.vector(t(varcor / attr(varcor, "stddev"))
                                 / attr(varcor, "stddev")))
    attr(varcor, "stddev") <- NULL
    attr(varcor, "correlation") <- NULL
    expect_equal_tol1(varcor, varcor0)

    # ranef
    ranef0 <- matrix(c(-0.1, -0.4,  0.5,  0.1,  0.3, -0.2,  0.9,  0.5,  0.0, -0.7,
                       -0.7,  0.0, -0.8,  1.2, -0.7,
                        0.8,  0.2, -0.6,  0.2, -0.8, -0.1, -0.6, -0.5, -0.2,  0.3,
                        1.0,  0.1,  0.7, -1.1,  0.7,
                        1.5, -0.1, -0.6,  0.5, -1.2, -0.5,  0.2, -0.3, -0.3, -0.2,
                        1.1,  0.1,  0.2, -0.7,  0.5,
                       -0.3,  0.3, -0.3, -0.2,  0.1,  0.3, -0.8, -0.3,  0.1,  0.6,
                        0.2,  0.0,  0.6, -0.7,  0.4),
                     15, 4)
    rownames(ranef0) <- as.character(1:15)
    colnames(ranef0) <- c("(Intercept)", "period2", "period3", "period4")
    expect_equal_tol1(as.matrix(ranef(model)[["herd"]]), ranef0)
})

test_that("succeeds on cbpp in parallel", {
    suppressWarnings({
        model <- mhglm(cbind(incidence, size - incidence) ~ period + (period | herd),
                       data = lme4::cbpp, family = binomial)
        model_p <- mhglm(cbind(incidence, size - incidence) ~ period + (period | herd),
                         data = lme4::cbpp, family = binomial,
                         control = list(parallel = TRUE))
    })

    expect_equal(fixef(model_p), fixef(model))
    expect_equal(vcov(model_p), vcov(model))
    expect_equal(VarCorr(model_p), VarCorr(model))
    expect_equal(ranef(model_p), ranef(model))
})

test_that("formula parameter takes character or variable input", {
    formula_var <- Reaction ~ Days + (Days | Subject)
    formula_char_vac <- "Reaction ~ Days + (Days | Subject)"

    model <- mhglm(Reaction ~ Days + (Days | Subject), data=lme4::sleepstudy)
    model_char <- mhglm("Reaction ~ Days + (Days | Subject)", data=lme4::sleepstudy)
    model_var <- mhglm(formula_var, data=lme4::sleepstudy)
    model_var_char <- mhglm(formula_char_vac, data=lme4::sleepstudy)

    expect_equal(fixef(model), fixef(model_char))
    expect_equal(fixef(model), fixef(model_var))
    expect_equal(fixef(model), fixef(model_var_char))

    expect_equal(ranef(model), ranef(model_char))
    expect_equal(ranef(model), ranef(model_var))
    expect_equal(ranef(model), ranef(model_var_char))
})

test_that("fit model with no data parameter", {
    Reaction <- lme4::sleepstudy$Reaction
    Days <- lme4::sleepstudy$Days
    Subject <- lme4::sleepstudy$Subject

    model <- mhglm(Reaction ~ Days + (Days | Subject), data=lme4::sleepstudy)
    model_no_data <- mhglm(Reaction ~ Days + (Days | Subject))

    expect_equal(fixef(model), fixef(model_no_data))
    expect_equal(ranef(model), ranef(model_no_data))
})

test_that("success using diagonal covariance", {
    model <- mhglm(Reaction ~ Days + (Days || Subject), data = lme4::sleepstudy,
                   control = mhglm.control(diagcov = TRUE))
    varcor <- VarCorr(model)[["Subject"]]
    expect_equal(varcor[1, 2], 0)
})
patperry/r-mbest documentation built on May 24, 2019, 8:20 p.m.