tests/testthat/test-tinyVAST.R

if (FALSE) {
if (requireNamespace("tinyVAST", quietly = TRUE)) {
  library("tinyVAST", warn.conflicts = FALSE)
  TOL <- 1e-5

  get_sdmTMB_pars <- function(x) {
    c(
      as.list(x$sd_report, "Estimate"),
      as.list(x$sd_report, "Estimate", report = TRUE)
    )
  }

  test_that("tinyVAST/sdmTMB Tweedie spatiotemporal IID models and index area integration match", {
    skip_on_cran()
    mesh <- make_mesh(pcod, c("X", "Y"), cutoff = 18)
    system.time({
      fit_sd <- sdmTMB(
        density ~ 0 + as.factor(year),
        data = pcod,
        mesh = mesh,
        family = tweedie(),
        time = "year",
        control = sdmTMBcontrol(multiphase = FALSE)
      )
    })
    ps <- get_sdmTMB_pars(fit_sd)

    system.time({
      fit_tv <- tinyVAST(
        density ~ 0 + factor(year),
        dsem = "",
        sem = "",
        data = pcod,
        family = tweedie(),
        time_column = "year",
        space_columns = c("X", "Y"),
        spatial_graph = mesh$mesh,
        control = tinyVASTcontrol(newton_loops = 1)
      )
    })
    pt <- as.list(fit_tv$sdrep, "Estimate")

    expect_equal(pt$alpha_j, ps$b_j, tolerance = TOL)
    expect_equal(pt$log_sigma[2], ps$thetaf, tolerance = TOL)
    expect_equal(pt$log_sigma[1], ps$ln_phi, tolerance = TOL)
    expect_equal(pt$beta_z, ps$sigma_E[1, 1], tolerance = TOL)
    expect_equal(pt$theta_z, ps$sigma_O[1, 1], tolerance = TOL)
    expect_equal(pt$log_kappa, ps$ln_kappa[1, 1], tolerance = TOL)

    g <- replicate_df(qcs_grid, "year", unique(pcod$year))
    p <- predict(fit_sd, newdata = g, return_tmb_object = TRUE)

    system.time({
      is <- get_index(p, bias_correct = TRUE)
    })

    # system.time({
    # is2 <- lapply(unique(g$year), \(x) {
    #   pp <- predict(fit_sd, newdata = subset(g, year == x), return_tmb_object = TRUE)
    #   get_index(pp, bias_correct = TRUE)
    # })
    # is2 <- do.call(rbind, is2)
    # })

    system.time({
      g$var <- "response"
      it <- lapply(unique(g$year), \(x)
      integrate_output(fit_tv, newdata = subset(g, year == x), apply.epsilon = TRUE))
      it <- do.call(rbind, it) |> as.data.frame()
    })

    expect_equal(it$`Est. (bias.correct)`, is$est, tolerance = TOL)
  })

  fit_sd <- sdmTMB(
    density ~ 0 + as.factor(year),
    data = pcod,
    family = delta_lognormal(type = "poisson-link"),
    spatial = "off",
    spatiotemporal = "off"
  )
  ps <- get_pars(fit_sd)

  suppressWarnings({
    fit_tv <- tinyVAST(
      density ~ 0 + factor(year),
      delta_options = list(delta_formula = ~ 0 + factor(year)),
      data = pcod,
      family = delta_lognormal(type = "poisson-link"),
      control = tinyVASTcontrol(newton_loops = 1)
    )
  })
  pt <- as.list(fit_tv$sdrep, "Estimate")

  compare_models_delta <- function(mtv, msd) {
    ps <- get_pars(msd)
    pt <- as.list(mtv$sdrep, "Estimate")
    expect_equal(logLik(msd), logLik(mtv), ignore_attr = TRUE)
    expect_equal(pt$alpha_j, ps$b_j, tolerance = TOL)
    expect_equal(pt$alpha2_j, ps$b_j2, tolerance = TOL)
  }

  test_that("tinyVAST/sdmTMB delta_lognormal(type = 'poisson-link') models match", {
    skip_on_cran()
    compare_models_delta(fit_tv, fit_sd)
    expect_equal(pt$log_sigma, ps$ln_phi[2], tolerance = TOL)
  })

  test_that("tinyVAST/sdmTMB delta_gamma() models match", {
    skip_on_cran()
    fit_sd <- update(fit_sd, family = delta_gamma())
    fit_tv <- update(fit_tv, family = delta_gamma())
    compare_models_delta(fit_tv, fit_sd)
    ps <- get_pars(fit_sd)
    pt <- as.list(fit_tv$sdrep, "Estimate")
    expect_equal(1 / (exp(pt$log_sigma))^2, exp(ps$ln_phi[2]), tolerance = TOL)
  })

  test_that("tinyVAST/sdmTMB delta_gamma(type = 'poisson-link')) models match", {
    skip_on_cran()
    fit_sd <- update(fit_sd, family = delta_gamma(type = "poisson-link"))
    fit_tv <- update(fit_tv, family = delta_gamma(type = "poisson-link"))
    compare_models_delta(fit_tv, fit_sd)
    ps <- get_pars(fit_sd)
    pt <- as.list(fit_tv$sdrep, "Estimate")
    expect_equal(1 / (exp(pt$log_sigma))^2, exp(ps$ln_phi[2]), tolerance = TOL)
  })
}
}

Try the sdmTMB package in your browser

Any scripts or data that you put into this service are public.

sdmTMB documentation built on June 22, 2024, 10:48 a.m.