tests/testthat/test-rpost_gev.R

In revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis

#context("GEV: rpost_rcpp vs rpost")

# We check that the values simulated using rpost() and rpost_rcpp() are
# (close enough to) identical when they are called using the same data,
# the same prior and starting from the same random number seed.

# Set a tolerance for the comparison of the simulated values
my_tol <- 1e-5

gev_test <- function(seed = 47, prior, n = 5, rotate = TRUE, trans = "none",
                     use_phi_map = FALSE, data = portpirie, nrep = 2, ...){
  if (prior == "user") {
    prior_rfn <- set_prior(prior = "norm", model = "gev", ...)
    ptr_gev_norm <- create_prior_xptr("gev_norm")
    prior_cfn <- set_prior(prior = ptr_gev_norm, model = "gev",
                           mean = c(0,0,0),
                           icov = solve(diag(c(10000, 10000, 100))))
  } else {
    prior_rfn <- set_prior(prior = prior, model = "gev", ...)
    prior_cfn <- set_prior(prior = prior, model = "gev", ...)
  }
  set.seed(seed)
  res1 <- rpost(n = n, model = "gev", prior = prior_rfn,
                     data = data, rotate = rotate, trans = trans,
                     use_phi_map = use_phi_map, nrep = nrep)
  set.seed(seed)
  res2 <- rpost_rcpp(n = n, model = "gev", prior = prior_cfn,
                     data = data, rotate = rotate, trans = trans,
                     use_phi_map = use_phi_map, nrep = nrep)
  return(list(sim1 = as.numeric(res1$sim_vals),
              sim2 = as.numeric(res2$sim_vals),
              data_rep1 = as.numeric(res1$data_rep),
              data_rep2 = as.numeric(res2$data_rep)))
}

test_function <- function(x, test_string) {
  testthat::test_that(test_string, {
    testthat::expect_equal(x$sim1, x$sim2, tolerance = my_tol)
    testthat::expect_equal(x$data_rep1, x$data_rep2, tolerance = my_tol)
  })
}

# Slow, belt-and-braces
#rotate_vals <- c(FALSE, TRUE)
#trans_vals <- c("none", "BC")
#use_phi_map_vals <- c(FALSE, TRUE)

# Faster, sufficient
rotate_vals <- TRUE
trans_vals <- "BC"
use_phi_map_vals <- TRUE

for (rotate in rotate_vals) {
  for (trans in trans_vals) {
    for (use_phi_map in use_phi_map_vals) {
      test_string <- paste("rotate =", rotate, "trans =", trans,
                            "use_phi_map =", use_phi_map)
      x <- gev_test(prior = "flat", min_xi = -1,
                    rotate = rotate, trans = trans, use_phi_map = use_phi_map)
      test_function(x, test_string)
      x <- gev_test(prior = "norm", mean = c(0,0,0),
                    cov = diag(c(10000, 10000, 100)),
                    rotate = rotate, trans = trans, use_phi_map = use_phi_map)
      test_function(x, test_string)
      x <- gev_test(prior = "loglognorm", mean = c(0,0,0),
                    cov = diag(c(10000, 10000, 100)),
                    rotate = rotate, trans = trans, use_phi_map = use_phi_map)
      test_function(x, test_string)
      x <- gev_test(prior = "mdi",
                    rotate = rotate, trans = trans, use_phi_map = use_phi_map)
      test_function(x, test_string)
      x <- gev_test(prior = "beta",
                    rotate = rotate, trans = trans, use_phi_map = use_phi_map)
      test_function(x, test_string)
      x <- gev_test(prior = "user", mean = c(0,0,0),
                    cov = diag(c(10000, 10000, 100)),
                    rotate = rotate, trans = trans, use_phi_map = use_phi_map)
      test_function(x, test_string)
      x <- gev_test(prior = "prob", quant = c(85, 88, 95),
                    alpha = c(4, 2.5, 2.25, 0.25), data = oxford,
                    rotate = rotate, trans = trans, use_phi_map = use_phi_map)
      test_function(x, test_string)
    }
  }
}