tests/testthat/test-mediate.R

In intmed: Mediation Analysis using Interventional Effects

context("Testing the mediate function")

set.seed(1234)



###################################################
# sim_data example - 1 mediator
###################################################

data("sim_data", package = "intmed")
tol = 0.006
tol2 = 0.009
med_res <- mediate(y = "y", med = c("m"), treat = "x", ymodel = "regression", mmodel = c("regression"), treat_lv = 1, control_lv = 0, incint = TRUE, inc_mmint = FALSE, data = sim_data, sim = 5000, digits = 5, HTML_report = FALSE, cores = 2)

test_that("Test using sim_data - 1 mediator", {
  skip_on_cran()
  skip_on_travis()
  expect_equal(mean(med_res$combined$direct), 0.523, tol)
  expect_equal(mean(med_res$combined$indirect[[1]]), 0.207, tol)
  expect_equal(mean(med_res$combined$total), 0.730, tol)
  expect_equal(median(med_res$combined$prop[[1]]), 0.283, tol2)
})


###################################################
# substance example - 2 mediators
###################################################
#tol = 0.006
#tol2 = 0.0015

#data("substance", package = "intmed")
#med_res <- intmed::mediate(y = "sub_misuse", med = c("dev_peer","sub_exp"), treat = "fam_int", c = c("conflict","gender"), ymodel = "logistic regression", mmodel = c("logistic regression","logistic regression"), treat_lv = 1, control_lv = 0, data = substance, sim = 1000, digits = 5, HTML_report = FALSE, cores = 2)

#test_that("Test using substance data - 2 mediators", {
#  skip_on_cran()
#  skip_on_travis()
#  expect_equal(mean(med_res$combined$direct), -0.05318 , tol)
#  expect_equal(mean(med_res$combined$indirect[[1]]), -0.01811, tol)
#  expect_equal(mean(med_res$combined$indirect[[2]]), -0.00622, tol)
#  expect_equal(mean(med_res$combined$total), -0.07670, tol)
#  expect_equal(median(med_res$combined$prop[[1]]), 0.22564, tol2)
#  expect_equal(median(med_res$combined$prop[[2]]), 0.07174, tol2)
#})

###################################################
# simulating data - 3 mediator
###################################################

#tol = 0.1

#n = 10000
#x = runif(n)
#x = ifelse(x < 0.5, 0, 1)

#c = rnorm(n,0,0.1)

#m1 = 0.2*x + rnorm(n, 0, 0.1) + 0.1*c
#m2 = 0.3*x + 0.9*m1 + rnorm(n, 0, 0.1) + 0.2*c
#m3 = 0.4*x + m1 + 1.1*m2 + rnorm(n, 0, 0.1) + 0.3*c

#y = 0.8*x + 0.5*m1 + 0.6*m2 + 0.7*m3 + rnorm(n, 0, 0.1) + 2 + 0.4*c

#simdata <- data.frame(x,m1,m2,m3,y,c)
#med_res <- mediate(y = "y", med = c("m1","m2","m3"), treat = "x", c = "c", ymodel = "regression", mmodel = c("regression", "regression", "regression"), treat_lv = 1, control_lv = 0, incint = NULL, inc_mmint = FALSE, data = simdata, sim = 5000, digits = 5, cores = 2)

#test_that("Test using simulated data - 3 mediators", {
#  skip_on_cran()
#  skip_on_travis()
#  expect_equal(mean(med_res$combined$direct), 0.8, tol)
#  expect_equal(mean(med_res$combined$indirect[[1]]), 0.1, tol)
#  expect_equal(mean(med_res$combined$indirect[[2]]), 0.288, tol)
#  expect_equal(mean(med_res$combined$indirect[[3]]), 0.7896, tol)
#  expect_equal(mean(med_res$combined$total), 1.9776, tol)
#})


###################################################
# simulating data with missing data - 3 mediators
###################################################

#simdata$rand = runif(nrow(simdata),0,1)
#simdata$y[simdata$rand < 0.1] <- NA
#simdata$rand = runif(nrow(simdata),0,1)
#simdata$x[simdata$rand < 0.05] <- NA
#simdata$rand = runif(nrow(simdata),0,1)
#simdata$m1[simdata$rand < 0.02] <- NA
#simdata$rand = runif(nrow(simdata),0,1)
#simdata$m2[simdata$rand < 0.01] <- NA
#simdata$rand = runif(nrow(simdata),0,1)
#simdata$m3[simdata$rand < 0.01] <- NA

#med_res <- mediate(y = "y", med = c("m1","m2","m3"), treat = "x", c = "c", ymodel = "regression", mmodel = c("regression", "regression", "regression"), treat_lv = 1, control_lv = 0, incint = NULL, inc_mmint = FALSE, data = simdata, sim = 1000, digits = 5)

#test_that("Test using simulated data with missing data - 3 mediators", {
#  skip_on_cran()
#  expect_equal(mean(med_res$combined$direct), 0.8, tol)
#  expect_equal(mean(med_res$combined$indirect[[1]]), 0.1, tol)
#  expect_equal(mean(med_res$combined$indirect[[2]]), 0.288, tol)
#  expect_equal(mean(med_res$combined$indirect[[3]]), 0.7896, tol)
#  expect_equal(mean(med_res$combined$total), 1.9776, tol)
#})