tests/testthat/OLDtest-5kernels.R

In haziqjamil/iprobit: Binary and Multinomial Probit Regression with I-priors

# context("Kernels and interactions")
#
# # Binary probit models ---------------------------------------------------------
#
# test_that("FBM kernel (binary)", {
#
#   dat <- gen_mixture(n = 10)
#
#   # FBM = 0.5
#   mod <- iprobit(y ~ ., dat, kernel = "FBM", silent = TRUE)
#   expect_equivalent(get_Hurst(mod), c(0.5, 0.5))
#   mod <- iprobit(dat$y, dat$X, kernel = "FBM", silent = TRUE,
#                  control = list(maxit = 2))
#   expect_equivalent(get_Hurst(mod), 0.5)
#
#   # FBM = 0.1, 0.9
#   mod <- iprobit(y ~ ., dat, kernel = c("FBM,0.1", "FBM,0.9"), silent = TRUE)
#   expect_equivalent(get_Hurst(mod), c(0.1, 0.9))
#
# })
#
# test_that("Pearson kernel (binary)", {
#
#   dat <- gen_mixture(n = 10)
#   mod <- iprobit(dat$y, dat$y, silent = TRUE)
#   expect_equivalent(get_kernel(mod), "Pearson")
#
# })
#
# test_that("Mixed kernel (binary)", {
#
#   dat <- gen_mixture(n = 10)
#   mod <- iprobit(y ~ ., dat, kernel = c("Canonical", "FBM"), silent = TRUE)
#   expect_equivalent(get_kernel(mod), c("Canonical", "FBM,0.5"))
#
# })
#
# test_that("Single lambda (binary)", {
#
#   dat <- gen_mixture(n = 10)
#   mod <- iprobit(dat$y, dat$X, silent = TRUE,
#                  control = list(alpha0 = 1, lambda0 = rep(1, 2)))
#   modf <- iprobit(y ~ ., dat, one.lam = TRUE, silent = TRUE,
#                   control = list(alpha0 = 1, lambda0 = rep(1, 2)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# test_that("Parsimonious interactions (binary)", {
#
#   dat <- gen_mixture(n = 100)
#   mod <- iprobit(dat$y, dat$X[, 1], dat$X[, 2], silent = TRUE,
#                  interactions = "1:2", parsm = TRUE,
#                  control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2)))
#   modf <- iprobit(y ~ . ^ 2, dat, silent = TRUE, parsm = TRUE,
#                   control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# test_that("Non-parsimonious interactions (binary)", {
#
#   dat <- gen_mixture(n = 100)
#   mod <- iprobit(dat$y, dat$X[, 1], dat$X[, 2], silent = TRUE,
#                  interactions = "1:2", parsm = FALSE,
#                  control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 3)))
#   modf <- iprobit(y ~ . ^ 2, dat, silent = TRUE, parsm = FALSE,
#                   control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 3)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# test_that("Squared terms (binary)", {
#
#   dat <- gen_mixture(n = 100)
#   mod <- iprobit(dat$y, dat$X[, 1], dat$X[, 1] ^ 2, silent = TRUE,
#                  control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2)))
#   modf <- iprobit(y ~ X1 + I(X1 ^ 2), dat, silent = TRUE,
#                   control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# # Multinomial probit models ----------------------------------------------------
#
# test_that("FBM kernel (multinomial)", {
#
#   dat <- gen_circle(n = 3, m = 3)
#
#   # FBM = 0.5
#   mod <- iprobit(y ~ ., dat, kernel = "FBM", silent = TRUE)
#   expect_equivalent(get_Hurst(mod), c(0.5, 0.5))
#   mod <- iprobit(dat$y, dat$X, kernel = "FBM", silent = TRUE,
#                  control = list(maxit = 2))
#   expect_equivalent(get_Hurst(mod), 0.5)
#
#   # FBM = 0.1, 0.9
#   mod <- iprobit(y ~ ., dat, kernel = c("FBM,0.1", "FBM,0.9"), silent = TRUE)
#   expect_equivalent(get_Hurst(mod), c(0.1, 0.9))
#
# })
#
# test_that("Pearson kernel (multinomial)", {
#
#   dat <- gen_circle(n = 3, m = 3)
#   mod <- iprobit(dat$y, dat$y, silent = TRUE)
#   # expect_equivalent(get_kernel(mod), "Pearson")
#
# })
#
# test_that("Mixed kernel (multinomial)", {
#
#   dat <- gen_circle(n = 3, m = 3)
#   mod <- iprobit(y ~ ., dat, kernel = c("Canonical", "FBM"), silent = TRUE)
#   expect_equivalent(get_kernel(mod), c("Canonical", "FBM,0.5"))
#
# })
#
# test_that("Single lambda (multinomial)", {
#
#   dat <- gen_circle(n = 3, m = 3)
#   mod <- iprobit(dat$y, dat$X, silent = TRUE,
#                  control = list(alpha0 = 1, lambda0 = rep(1, 2 * 3)))
#   modf <- iprobit(y ~ ., dat, one.lam = TRUE, silent = TRUE,
#                   control = list(alpha0 = 1, lambda0 = rep(1, 2 * 3)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# test_that("Parsimonious interactions (multinomial)", {
#
#   dat <- gen_mixture(n = 100, m = 3)
#   mod <- iprobit(dat$y, dat$X[, 1], dat$X[, 2], silent = TRUE,
#                  interactions = "1:2", parsm = TRUE,
#                  control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2 * 3)))
#   modf <- iprobit(y ~ . ^ 2, dat, silent = TRUE, parsm = TRUE,
#                   control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2 * 3)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# test_that("Non-parsimonious interactions (multinomial)", {
#
#   dat <- gen_mixture(n = 100, m = 3)
#   mod <- iprobit(dat$y, dat$X[, 1], dat$X[, 2], silent = TRUE,
#                  interactions = "1:2", parsm = FALSE,
#                  control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 3 * 3)))
#   modf <- iprobit(y ~ . ^ 2, dat, silent = TRUE, parsm = FALSE,
#                   control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 3 * 3)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# test_that("Squared terms (multinomial)", {
#
#   dat <- gen_mixture(n = 100, m = 3)
#   mod <- iprobit(dat$y, dat$X[, 1], dat$X[, 1] ^ 2, silent = TRUE,
#                  control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2)))
#   modf <- iprobit(y ~ X1 + I(X1 ^ 2), dat, silent = TRUE,
#                   control = list(maxit = 3, alpha0 = 1, lambda0 = rep(1, 2)))
#   expect_equivalent(coef(mod), coef(modf))
#
# })
#
# test_that("Lots of squared terms (multinomial)", {
#
#   data(iris)
#   mod <- iprobit(Species ~ . ^ 2, iris, silent = TRUE,
#                  control = list(maxit = 2))
#   expect_s3_class(mod, "iprobitMod")
#
# })