tests/testthat/test_w2n.R

In TheoMichelot/moveHMM: Animal Movement Modelling using Hidden Markov Models

context("w2n")

test_that("The output is in the right format", {
    nbStates <- 2
    nbCovs <- 2
    par <- c(0.001, 0.999, 0.5, 0.7, 1500.3, 0.001)
    parSize <- c(2, 1)
    bounds <- matrix(c(0, 1, 0, 1, 0, 1,
                       0, 1, 0, Inf, 0, Inf),
                     byrow = TRUE, ncol = 2)
    beta <- matrix(rnorm(6), ncol = 2, nrow = 3)
    delta <- c(0.6, 0.4)
    wpar <- n2w(par, bounds, beta, delta, nbStates, FALSE)
    p <- w2n(wpar, bounds, parSize, nbStates, nbCovs, FALSE, FALSE)

    expect_equal(length(p$stepPar), parSize[1]*nbStates)
    expect_equal(length(p$anglePar), parSize[2]*nbStates)
    expect_equal(dim(p$beta), dim(beta))
    expect_equal(length(p$delta), length(delta))
})

test_that("w2n and n2w are inverse", {
    nbStates <- 2
    nbCovs <- 2
    par <- c(0.001, 0.999, 0.7, 0.7, 1500.3, 0.001)
    parSize <- c(2, 1)
    bounds <- matrix(c(0, 1, 0, 1, 0, 1,
                       0, 1, -Inf, Inf, -Inf, Inf),
                     byrow = TRUE, ncol = 2)
    beta <- matrix(rnorm(6), ncol = 2, nrow = 3)
    delta <- c(0.6, 0.4)
    wpar <- n2w(par, bounds, beta, delta, nbStates, FALSE)
    p <- w2n(wpar, bounds, parSize, nbStates, nbCovs, FALSE, FALSE)

    expect_equal(c(t(p$stepPar)), par[1:(parSize[1]*nbStates)], tolerance = 1e-10)
    expect_equal(c(t(p$anglePar)), par[(parSize[1]*nbStates+1):(sum(parSize)*nbStates)], tolerance = 1e-10)
    expect_equal(p$beta, beta, tolerance = 1e-10)
    expect_equal(p$delta, delta, tolerance = 1e-10)
})