tests/testthat/test.c_random_mvnorm.R

In ashiklom/mvtraits: Fit multivariate models to partially missing trait data

context("C random mvnorm")

ss <- function(x, y) {
    sum((y - x)^2)
}

mu <- colMeans(iris[,-5])
Sigma <- cov(iris[,-5])

n <- 5000
m <- length(mu)

test_that('c_random_mvnorm (vector mu) produces reasonable values', {
  samp <- random_mvnorm(n, t(mu), Sigma)
  samp_mu <- colMeans(samp)
  expect_lt(ss(samp_mu, mu), 0.01)
  samp_cov <- cov(samp)
  expect_lt(ss(samp_cov, Sigma), 0.05)
})

test_that('c_random_mvnorm (matrix mu) produces reasonable values', {
  mu_mat <- matrix(mu, n, m, byrow = TRUE)
  samp2 <- mvtraits:::c_random_mvnorm(n, mu_mat, Sigma)
  samp2_mu <- colMeans(samp2)
  samp2_cov <- cov(samp2)
  expect_lt(ss(samp2_mu, mu), 0.01)
  expect_lt(ss(samp2_cov, Sigma), 0.05)
})