add tests for the samplers

tests/testthat/test-samplers.R

@@ -0,0 +1,97 @@
+# Tests for samplers
+
+# post_draws()
+test_that("post_draws() works for a GAM", {
+  expect_silent(drws1 <- post_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2))
+  expect_silent(drws2 <- post_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2))
+  expect_type(drws1, "double")
+  expect_true(is.matrix(drws1))
+  expect_identical(dim(drws1), c(20L, 37L))
+  expect_identical(drws1, drws2)
+})
+
+test_that("post_draws() works for a GAM", {
+  expect_silent(drws1 <- post_draws(m_gam, n = 20, method = "mh",
+    burnin = 100, thin = 2, t_df = 4, rw_scale = 0.3, seed = 2))
+  expect_silent(drws2 <- post_draws(m_gam, n = 20, method = "mh",
+    burnin = 100, thin = 2, t_df = 4, rw_scale = 0.3, seed = 2))
+  expect_type(drws1, "double")
+  expect_true(is.matrix(drws1))
+  expect_identical(dim(drws1), c(20L, 37L))
+  expect_identical(drws1, drws2)
+})
+
+test_that("post_draws() fails for INLA", {
+  expect_error(post_draws(m_gam, method = "inla"),
+    "'method = \"inla\"' is not yet implemented.")
+})
+
+test_that("post_draws() works for a GAM with mgcv mvn()", {
+  expect_silent(drws1 <- post_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2, mvn_method = "mgcv"))
+  expect_silent(drws2 <- post_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2, mvn_method = "mgcv"))
+  expect_type(drws1, "double")
+  expect_true(is.matrix(drws1))
+  expect_identical(dim(drws1), c(20L, 37L))
+  expect_identical(drws1, drws2)
+})
+
+# generate_draws()
+test_that("generate_draws() works for a GAM", {
+  expect_silent(drws1 <- generate_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2))
+  expect_silent(drws2 <- generate_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2))
+  expect_type(drws1, "double")
+  expect_true(is.matrix(drws1))
+  expect_identical(dim(drws1), c(20L, 37L))
+  expect_identical(drws1, drws2)
+})
+
+test_that("generate_draws() works for a GAM with MH", {
+  expect_silent(drws1 <- generate_draws(m_gam, n = 20, method = "mh",
+    burnin = 100, thin = 2, t_df = 4, rw_scale = 0.3, seed = 2))
+  expect_silent(drws2 <- generate_draws(m_gam, n = 20, method = "mh",
+    burnin = 100, thin = 2, t_df = 4, rw_scale = 0.3, seed = 2))
+  expect_type(drws1, "double")
+  expect_true(is.matrix(drws1))
+  expect_identical(dim(drws1), c(20L, 37L))
+  expect_identical(drws1, drws2)
+})
+
+test_that("generate_draws() fails for INLA", {
+  expect_error(generate_draws(m_gam, method = "inla"),
+    "'method = \"inla\"' is not yet implemented.")
+})
+
+test_that("generate_draws() works for a GAM with mgcv mvn()", {
+  expect_silent(drws1 <- generate_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2, mvn_method = "mgcv"))
+  expect_silent(drws2 <- generate_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2, mvn_method = "mgcv"))
+  expect_type(drws1, "double")
+  expect_true(is.matrix(drws1))
+  expect_identical(dim(drws1), c(20L, 37L))
+  expect_identical(drws1, drws2)
+})
+
+# user_draws()
+test_that("user_draws() works for a GAM", {
+  expect_silent(drws1 <- generate_draws(m_gam, n = 20, method = "gaussian",
+    seed = 2))
+  expect_silent(udrws <- user_draws(m_gam, draws = drws1))
+  expect_type(udrws, "double")
+  expect_true(is.matrix(udrws))
+  expect_identical(dim(udrws), c(20L, 37L))
+  expect_identical(udrws, drws1)
+})
+
+test_that("user_draws() fails for incorrect matrix of draws", {
+  expect_error(user_draws(m_gam, draws = drws1[, 1:10]),
+    "Supplied 'draws' doesn't match number of model coefficients.
+Number of model coefs: 37
+Number of columns in 'draws': 10")
+})