context("stepFit") source(system.file("tests/comparisons/localEst.R", package = "stepR")) source(system.file("tests/comparisons/localCost.R", package = "stepR")) # confidence intervals testConfInt <- function(s, b, x, n, tolerance = 1e-9) { # end vectors expect_identical(s$leftEnd, x[s$leftIndex]) expect_identical(s$rightEnd, x[s$rightIndex]) expect_identical(s$leftEndLeftBound, x[s$leftIndexLeftBound]) expect_identical(s$leftEndRightBound, x[s$leftIndexRightBound]) expect_identical(s$rightEndLeftBound, x[s$rightIndexLeftBound]) expect_identical(s$rightEndRightBound, x[s$rightIndexRightBound]) L <- s$leftIndexLeftBound[-1] R <- s$leftIndexRightBound[-1] K <- length(L) expect_identical(s$leftIndexLeftBound[1], 1L) expect_identical(s$leftIndexRightBound[1], 1L) # test L if (K > 0L) { expect_true(max(c(-Inf, b$lower[b$li >= L[K]])) <= min(c(Inf, b$upper[b$li >= L[K]])) + tolerance) expect_true(max(c(-Inf, b$lower[b$li >= L[K] - 1L])) > min(c(Inf, b$upper[b$li >= L[K] - 1L])) - tolerance) if (K > 1L) { for (i in (K - 1L):1L) { # L[i] can be reached from L[i + 1] - 1 expect_true(max(c(-Inf, b$lower[b$li >= L[i] & b$ri <= L[i + 1] - 1L])) <= min(c( Inf, b$upper[b$li >= L[i] & b$ri <= L[i + 1] - 1L])) + tolerance) # L[i] - 1 cannot be reached from L[i + 1] - 1 expect_true(max(c(-Inf, b$lower[b$li >= L[i] - 1L & b$ri <= L[i + 1] - 1L])) > min(c( Inf, b$upper[b$li >= L[i] - 1L & b$ri <= L[i + 1] - 1L])) - tolerance) } } expect_true(max(c(-Inf, b$lower[b$ri <= L[1] - 1L])) <= min(c(Inf, b$upper[b$ri <= L[1] - 1L])) + tolerance) } else { expect_true(max(b$lower) <= min(b$upper) + tolerance) } # test R if (K > 0L) { expect_true(max(c(-Inf, b$lower[b$ri <= R[1]])) > min(c(Inf, b$upper[b$ri <= R[1]])) - tolerance) expect_true(max(c(-Inf, b$lower[b$ri <= R[1] - 1L])) <= min(c(Inf, b$upper[b$ri <= R[1] - 1L])) + tolerance) if (K > 1L) { for (i in 2:K) { # R[i] cannot be reached expect_true(max(c(-Inf, b$lower[b$li >= R[i - 1] & b$ri <= R[i]])) > min(c( Inf, b$upper[b$li >= R[i - 1] & b$ri <= R[i]])) - tolerance) # R[i] - 1 can be reached expect_true(max(c(-Inf, b$lower[b$li >= R[i - 1] & b$ri <= R[i] - 1L])) <= min(c( Inf, b$upper[b$li >= R[i - 1] & b$ri <= R[i] - 1L])) + tolerance) } } expect_true(max(c(-Inf, b$lower[b$li >= R[K]])) <= min(c(Inf, b$upper[b$li >= R[K]])) + tolerance) } # test s$rightIndexLeftBound and s$rightIndexRightBound expect_identical(s$rightIndexLeftBound, c(s$leftIndexLeftBound[-1] - 1L, n)) expect_identical(s$rightIndexRightBound, c(s$leftIndexRightBound[-1] - 1L, n)) } # feasibility of the solution testFeasible <- function(s, b, tolerance = 1e-9) { for (i in 1:length(s$leftIndex)) { expect_true(max(c(-Inf, b$lower[b$li >= s$leftIndex[i] & b$ri <= s$rightIndex[i]])) <= s$value[i] + tolerance) expect_true(min(c( Inf, b$upper[b$li >= s$leftIndex[i] & b$ri <= s$rightIndex[i]])) >= s$value[i] - tolerance) } } # test function values testValues <- function(s, b, localEst, y, tolerance = 1e-9, ...) { for (i in 1:length(s$leftIndex)) { lower <- max(c(-Inf, b$lower[b$li >= s$leftIndex[i] & b$ri <= s$rightIndex[i]])) upper <- min(c( Inf, b$upper[b$li >= s$leftIndex[i] & b$ri <= s$rightIndex[i]])) test <- localEst(y, s$leftIndex[i], s$rightIndex[i], lower, upper, s$leftIndex, s$rightIndex, s$value, ...) expect_equal(s$value[i], localEst(y, s$leftIndex[i], s$rightIndex[i], lower, upper, s$leftIndex, s$rightIndex, s$value, ...), tolerance = tolerance) } } # test optimality and costs costSolution <- function(y, left, right, est, localCost, ...) { costs <- 0 for (i in 1:length(left)) { costs <- costs + localCost(y, left[i], right[i], est[i], left, right, est, ...) } costs } # also included in testOptimality, only useful if testOptimality is skipped due to its runtime testCosts <- function(s, y, localCost, tolerance = 1e-9, ...) { opt <- costSolution(y, s$leftIndex, s$rightIndex, s$value, localCost, ...) expect_equal(attr(s, "cost"), opt, tolerance = tolerance) } testOptimality <- function(s, b, y, localCost, localEst, tolerance = 1e-9, ...) { left <- s$leftIndexLeftBound est <- numeric(length(left)) opt <- costSolution(y, s$leftIndex, s$rightIndex, s$value, localCost, ...) expect_equal(attr(s, "cost"), opt, tolerance = tolerance) if (length(left) > 1L) { while (left[length(left)] <= s$leftIndexRightBound[length(left)]) { right <- c(left[-1] - 1L, length(y)) feas <- TRUE # check feasibility for (i in 1:length(left)) { lower <- max(c(-Inf, b$lower[b$li >= left[i] & b$ri <= right[i]])) upper <- min(c( Inf, b$upper[b$li >= left[i] & b$ri <= right[i]])) if (upper < lower) { feas <- FALSE break } else { est[i] <- localEst(y, left[i], right[i], lower, upper, left, right, est, ...) } } # compare cost of the solution if (feas && !identical(s$leftIndex, left)) { expect_true(costSolution(y, left, right, est, localCost, ...) >= opt - tolerance) } left[2] <- left[2] + 1L if (length(left) > 2L) { for (i in 2:(length(left) - 1L)) { if (left[i] > s$leftIndexRightBound[i]) { left[i + 1] <- left[i + 1] + 1L left[i] <- s$leftIndexLeftBound[i] } } } } } } # confidence band testBand <- function(s, b, x, tolerance = 1e-9) { band <- confband(s) expect_identical(band$x, x) leftconst <- s$leftIndexRightBound rightconst <- s$rightIndexLeftBound for (i in 1:length(leftconst)) { if (rightconst[i] >= leftconst[i]) { lower <- max(c(-Inf, b$lower[b$li >= leftconst[i] & b$ri <= rightconst[i]])) upper <- min(c( Inf, b$upper[b$li >= leftconst[i] & b$ri <= rightconst[i]])) for (j in leftconst[i]:rightconst[i]) { expect_true((band$lower[j] == -Inf && lower == -Inf) || all.equal(lower, band$lower[j], tolerance = tolerance)) expect_true((band$upper[j] == Inf && upper == Inf) || all.equal(upper, band$upper[j], tolerance = tolerance)) } } } if (length(s$leftIndexLeftBound[-1]) > 0L) { for (i in 1:length(s$leftIndexLeftBound[-1])) { if (s$rightIndexRightBound[i] >= s$leftIndexLeftBound[i + 1]) { for (j in s$leftIndexLeftBound[i + 1]:s$rightIndexRightBound[i]) { lower <- min(max(c(-Inf, b$lower[b$li >= leftconst[i] & b$ri <= j])), max(c(-Inf, b$lower[b$li >= j & b$ri <= rightconst[i + 1]]))) upper <- max(min(c(Inf, b$upper[b$li >= leftconst[i] & b$ri <= j])), min(c(Inf, b$upper[b$li >= j & b$ri <= rightconst[i + 1]]))) expect_true((band$lower[j] == -Inf && lower == -Inf) || all.equal(lower, band$lower[j], tolerance = tolerance)) expect_true((band$upper[j] == Inf && upper == Inf) || all.equal(upper, band$upper[j], tolerance = tolerance)) } } } } } test_that("argument y is tested", { testn <- 100L testy <- rnorm(testn) testq <- rep(4, testn) expect_error(stepFit(q = testq)) expect_error(stepFit(numeric(0), q = testq)) expect_identical(stepFit(testy, q = testq), stepFit(testy, x = 1:testn, x0 = 0, family = "gauss", intervalSystem = "all", lengths = 1:testn, q = testq, confband = FALSE, jumpint = FALSE, sd = sdrobnorm(testy))) expect_error(stepFit(as.integer(testy), q = testq)) expect_error(stepFit(c(testy, "s"), q = testq)) expect_error(stepFit(c(rnorm(10), NA), q = testq)) s <- stepFit(testy, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument x works and is tested", { testy <- c(rnorm(40, 0), rnorm(5, 10), rnorm(5, 20), rnorm(5, 30), rnorm(40, 40)) testn <- length(testy) testq <- rep(4, testn) expect_error(stepFit(y = testy, x = 1:10, q = testq)) expect_error(stepFit(y = testy, x = c(1:214, "s"), q = testq)) expect_error(stepFit(y = testy, x = 215:1, q = testq)) expect_error(stepFit(y = testy, x = c(1:214, NA), q = testq)) expect_error(stepFit(y = testy, x = c(1:214, Inf), q = testq)) testx <- 1:testn expect_identical(stepFit(testy, x = testx, q = testq, sd = 1, confband = TRUE), stepFit(testy, q = testq, sd = 1, confband = TRUE)) testx <- 1:testn / 1e4 s <- stepFit(testy, x = testx, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testx <- testn:1 / -4.5e2 s <- stepFit(testy, x = testx, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testx <- c(0, 1.23, 3:95) s <- stepFit(testy, x = testx, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument x0 works and is tested", { testy <- c(rnorm(20, 0), rnorm(5, 10), rnorm(5, 20), rnorm(5, 30), rnorm(20, 40)) testn <- length(testy) testq <- rep(5, testn) expect_error(stepFit(y = testy, x0 = 2, q = testq)) expect_error(stepFit(y = testy, x0 = "0", q = testq)) expect_error(stepFit(y = testy, x0 = Inf, q = testq)) expect_error(stepFit(y = testy, x0 = c(0, 0.5), q = testq)) expect_identical(stepFit(testy, x0 = 0, q = testq, sd = 1, confband = TRUE), stepFit(testy, q = testq, sd = 1, confband = TRUE)) testx <- 1:testn / 1e4 s <- stepFit(testy, x = testx, x0 = -1, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument family is tested and works in the default case", { # other families below testn <- 53L testy <- c(rnorm(30, 1, 0.23), rnorm(testn - 30, -1, 0.34)) testq <- 53:1 / 20 expect_error(stepFit(testy, family = "", q = testq)) expect_error(stepFit(testy, family = c("gauss", "hsmuce"), q = testq)) expect_identical(stepFit(testy, q = testq), stepFit(testy, q = testq, family = "gauss")) testx <- 1:testn s <- stepFit(testy, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument intervalSystem is tested and works", { testn <- 23L testy <- c(rnorm(10, 1, 0.23), rnorm(testn - 10, -1, 0.34)) testq <- 23:1 / 5 expect_error(stepFit(testy, intervalSystem = "", q = testq)) expect_error(stepFit(testy, intervalSystem = "dya", q = testq)) expect_error(stepFit(testy, intervalSystem = "dyalen", q = testq)) expect_error(stepFit(testy, intervalSystem = "dyapar", q = testq)) expect_error(stepFit(testy, intervalSystem = c("all", "dyaLen"), q = testq)) expect_identical(stepFit(testy, q = testq), stepFit(testy, q = testq, intervalSystem = "all")) testq <- 5:1 / 5 testx <- 1:testn s <- stepFit(testy, intervalSystem = "dyaLen", q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, intervalSystem = "dyaLen", q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testq <- 5:1 / 5 testx <- 1:testn s <- stepFit(testy, intervalSystem = "dyaPar", q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, intervalSystem = "dyaPar", q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument lengths is tested and works", { testn <- 36L testy <- c(rnorm(12, 1, 0.23), rnorm(12, -1, 0.23), rnorm(12, 1, 0.23)) testq <- 36:1 / 8 expect_error(stepFit(testy, lengths = "s", q = testq)) expect_error(stepFit(testy, lengths = c(1:10, NA), q = testq)) expect_error(stepFit(testy, lengths = c(1:10, Inf), q = testq)) expect_error(stepFit(testy, lengths = 0:10, q = testq)) expect_error(stepFit(testy, lengths = -1L, q = testq)) expect_error(stepFit(testy, lengths = 38L, q = testq)) expect_warning(ret <- stepFit(testy, lengths = c(1:10, 10), q = testq)) expect_identical(ret, stepFit(testy, lengths = c(1:10), q = testq)) expect_identical(stepFit(testy, lengths = c(10:1), q = testq), stepFit(testy, lengths = c(1:10), q = testq)) expect_identical(stepFit(testy, lengths = c(1:10 + 0.5), q = testq), stepFit(testy, lengths = c(1:10), q = testq)) testq <- 2 expect_error(stepFit(testy, intervalSystem = "dyaLen", lengths = 3L, q = testq)) expect_error(stepFit(testy, intervalSystem = "dyaLen", lengths = 64L, q = testq)) expect_error(stepFit(testy, intervalSystem = "dyaPar", lengths = 3L, q = testq)) expect_error(stepFit(testy, intervalSystem = "dyaPar", lengths = 64L, q = testq)) expect_equal(stepFit(testy, lengths = 2^(0:5), q = testq), stepFit(testy, intervalSystem = "dyaLen", q = testq)) testq <- 9:1 testlengths = c(2, 5:10, 35:36) testx <- 1:testn s <- stepFit(testy, lengths = testlengths, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, lengths = testlengths, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testq <- 3:1 / 5 testlengths = c(2, 4, 32) testx <- 1:testn s <- stepFit(testy, lengths = testlengths, intervalSystem = "dyaLen", q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, intervalSystem = "dyaLen", lengths = testlengths, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testq <- 2:1 / 5 testlengths = c(1, 32) testx <- 1:testn s <- stepFit(testy, lengths = testlengths, intervalSystem = "dyaPar", q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, intervalSystem = "dyaPar", lengths = testlengths, q = testq, sd = 1) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument q is tested and works if it is given", { testn <- 36L testy <- c(rnorm(12, 1, 0.23), rnorm(12, -1, 0.23), rnorm(12, 1, 0.23)) teststat <- monteCarloSimulation(36L) expect_warning(ret <- stepFit(testy)) expect_identical(ret, stepFit(testy, alpha = 0.5)) expect_warning(ret <- stepFit(testy, stat = teststat)) expect_identical(ret, stepFit(testy, stat = teststat, alpha = 0.5)) expect_error(stepFit(testy, q = "s", alpha = 0.1, stat = teststat)) expect_error(stepFit(testy, q = c(rep(1, 35), "s"), alpha = 0.1, stat = teststat)) expect_error(stepFit(testy, q = rep(1, 37), alpha = 0.1, stat = teststat)) expect_error(stepFit(testy, q = rep(1, 33), alpha = 0.1, stat = teststat)) testq <- 1:36 attr(testq, "n") <- "s" expect_error(stepFit(testy, q = testq, alpha = 0.1, stat = teststat)) attr(testq, "n") <- 35L expect_error(stepFit(testy, q = testq, alpha = 0.1, stat = teststat)) expect_identical(stepFit(testy, q = 3, intervalSystem = "all", lengths = c(1:3, 8:23)), stepFit(testy, intervalSystem = "all", lengths = c(1:3, 8:23), q = critVal(q = 3, intervalSystem = "all", lengths = c(1:3, 8:23), n = testn))) expect_identical(stepFit(testy, q = 3, intervalSystem = "dyaLen", lengths = c(1, 4, 8, 32), nq = 45L), stepFit(testy, intervalSystem = "dyaLen", lengths = c(1, 4, 8, 32), q = critVal(q = 3, intervalSystem = "dyaLen", lengths = c(1, 4, 8, 32), n = testn, nq = 45L))) expect_identical(stepFit(testy, q = 3, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32)), stepFit(testy, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), q = critVal(q = 3, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), n = testn))) testq <- 1:45 attr(testq, "n") <- 45L expect_identical(stepFit(testy, q = testq), stepFit(testy, q = 1:36)) expect_identical(stepFit(testy, q = 1:36, lengths = 3:23), stepFit(testy, q = 3:23, lengths = 3:23)) expect_identical(stepFit(testy, q = testq, lengths = 3:23), stepFit(testy, q = 3:23, lengths = 3:23)) expect_identical(stepFit(testy, q = 1:36, intervalSystem = "dyaLen"), stepFit(testy, q = 2^(0:5), intervalSystem = "dyaLen")) expect_identical(stepFit(testy, q = testq, intervalSystem = "dyaLen"), stepFit(testy, q = 2^(0:5), intervalSystem = "dyaLen")) testq <- 2^(0:6) attr(testq, "n") <- 64L expect_identical(stepFit(testy, q = testq, intervalSystem = "dyaPar"), stepFit(testy, q = 2^(0:5), intervalSystem = "dyaPar")) }) test_that("argument q is computed correctly", { testn <- 36L testy <- c(rnorm(12, 1, 0.23), rnorm(12, -1, 0.23), rnorm(12, 1, 0.23)) teststat <- monteCarloSimulation(n = testn, r = 100L) ret <- stepFit(testy, alpha = 0.1, stat = teststat) expect_identical(ret, stepFit(testy, alpha = 0.1, stat = teststat, nq = testn, family = "gauss", intervalSystem = "all", lengths = 1:testn, penalty = "sqrt")) expect_identical(ret, stepFit(testy, q = critVal(alpha = 0.1, stat = teststat, n = testn))) expect_error(stepFit(testy, alpha = "s", stat = teststat)) expect_error(stepFit(testy, alpha = 0, stat = teststat)) expect_identical(stepFit(testy, alpha = 0.075, stat = teststat), stepFit(testy, q = critVal(alpha = 0.075, stat = teststat, n = testn))) expect_error(stepFit(testy, alpha = 0.1, stat = teststat, n = testn)) expect_error(stepFit(testy, alpha = 0.1, nq = "s", stat = teststat)) expect_error(stepFit(testy, q = 1, nq = "s", stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, nq = Inf, stat = teststat)) expect_error(stepFit(testy, q = 1, nq = Inf, stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, nq = 8L, stat = teststat)) expect_error(stepFit(testy, q = 1, nq = 8L, stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, penalty = "", stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, penalty = "ads", stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, penalty = "weights", weights = NA, stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, penalty = "weights", weights = rep(1 / 8, 8), stat = teststat)) expect_identical(stepFit(testy, alpha = 0.1, penalty = "weights", stat = teststat), stepFit(testy, q = critVal(alpha = 0.1, penalty = "weights", weights = rep(1 / 36, 36), stat = teststat, n = testn))) expect_identical(stepFit(testy, alpha = 0.1, penalty = "weights", weights = rep(1, 36), stat = teststat), stepFit(testy, q = critVal(alpha = 0.1, penalty = "weights", weights = rep(1 / 36, 36), stat = teststat, n = testn))) teststat <- monteCarloSimulation(n = 36L, r = 100L) expect_identical(stepFit(testy, alpha = 0.14, penalty = "log", stat = teststat, intervalSystem = "all", lengths = c(1, 5, 8, 23)), stepFit(testy, intervalSystem = "all", lengths = c(1, 5, 8, 23), q = critVal(alpha = 0.14, intervalSystem = "all", lengths = c(1, 5, 8, 23), penalty = "log", stat = teststat, n = testn))) teststat <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "dyaLen") expect_identical(stepFit(testy, alpha = 0.034, penalty = "none", stat = teststat, intervalSystem = "dyaLen", lengths = c(1, 4, 8, 32)), stepFit(testy, intervalSystem = "dyaLen", lengths = c(1, 4, 8, 32), q = critVal(alpha = 0.034, intervalSystem = "dyaLen", lengths = c(1, 4, 8, 32), penalty = "none", stat = teststat, n = testn))) teststat <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "dyaPar") expect_identical(stepFit(testy, alpha = 0.56, penalty = "weights", weights = c(0.2, 0.3, 0.4, 0.1), stat = teststat, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32)), stepFit(testy, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), q = critVal(alpha = 0.56, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), penalty = "weights", weights = c(0.2, 0.3, 0.4, 0.1), stat = teststat, n = testn))) teststat <- matrix(abs(rnorm(1e4 * testn)), testn, 1e4) expect_error(stepFit(testy, alpha = 0.1, stat = teststat)) teststat <- monteCarloSimulation(n = 37L, r = 100L) expect_identical(stepFit(testy, alpha = 0.05, stat = teststat, nq = 37L), stepFit(testy, q = critVal(alpha = 0.05, stat = teststat, n = testn, nq = 37L))) teststat <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "dyaPar") expect_identical(stepFit(testy, alpha = 0.014, stat = teststat, nq = 100L, intervalSystem = "dyaPar", lengths = 2^c(0:3, 5)), stepFit(testy, intervalSystem = "dyaPar", lengths = 2^c(0:3, 5), q = critVal(alpha = 0.014, intervalSystem = "dyaPar", lengths = 2^c(0:3, 5), stat = teststat, n = testn, nq = 100L))) teststat <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "all") expect_identical(stepFit(testy, alpha = 0.014, stat = teststat, nq = 100L, intervalSystem = "all", lengths = 1L), stepFit(testy, intervalSystem = "all", lengths = 1L, q = critVal(alpha = 0.014, intervalSystem = "all", lengths = 1, stat = teststat, n = testn, nq = 100L))) expect_error(stepFit(testy, alpha = 0.1, output = "vector", stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, data = 1, stat = teststat)) teststat <- monteCarloSimulation(n = 36L, r = 100L, output = "maximum") expect_error(stepFit(testy, alpha = 0.1, penalty = "weights", stat = teststat)) teststatmatrix <- monteCarloSimulation(n = 36L, r = 100L) teststatvector <- monteCarloSimulation(n = 36L, r = 100L, output = "maximum", penalty = "sqrt") expect_identical(stepFit(testy, alpha = 0.05, stat = teststatvector), stepFit(testy, q = critVal(alpha = 0.05, stat = teststatmatrix, n = testn))) teststatvector <- monteCarloSimulation(n = 36L, r = 100L, output = "maximum", penalty = "log", lengths = c(1, 3, 5, 8, 9, 12)) expect_identical(stepFit(testy, alpha = 0.15, stat = teststatvector, nq = 100L, lengths = c(1, 3, 5, 8, 9, 12), penalty = "log"), stepFit(testy, lengths = c(1, 3, 5, 8, 9, 12), q = critVal(alpha = 0.15, stat = teststatmatrix, n = testn, nq = 100L, lengths = c(1, 3, 5, 8, 9, 12), penalty = "log"))) teststatmatrix <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "dyaLen") teststatvector <- monteCarloSimulation(n = 36L, r = 100L, output = "maximum", intervalSystem = "dyaLen", penalty = "sqrt", lengths = c(1, 2, 8, 16)) expect_identical(stepFit(testy, alpha = 0.05, stat = teststatvector, nq = 2^7L, intervalSystem = "dyaLen", lengths = c(1, 2, 8, 16), penalty = "sqrt"), stepFit(testy, intervalSystem = "dyaLen", lengths = c(1, 2, 8, 16), q = critVal(alpha = 0.05, stat = teststatmatrix, n = testn, nq = 2^7L, intervalSystem = "dyaLen", lengths = c(1, 2, 8, 16), penalty = "sqrt"))) teststatmatrix <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "dyaPar") teststatvector <- monteCarloSimulation(n = 36L, r = 100L, output = "maximum", intervalSystem = "dyaPar", penalty = "sqrt", lengths = c(1, 2, 8, 16, 32)) expect_identical(stepFit(testy, alpha = 0.122, stat = teststatvector, nq = 2^9L, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), penalty = "sqrt"), stepFit(testy, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), q = critVal(alpha = 0.122, stat = teststatmatrix, n = testn, nq = 2^9L, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), penalty = "sqrt"))) expect_identical(stepFit(testy, alpha = 0.122, stat = teststatvector, nq = 2^9L, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), penalty = "sqrt"), stepFit(testy, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), q = critVal(alpha = 0.122, stat = teststatvector, n = testn, nq = 2^9L, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), penalty = "sqrt", output = "vector"))) expect_identical(stepFit(testy, alpha = 0.122, stat = teststatvector, nq = 2^9L, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), penalty = "sqrt"), stepFit(testy, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), q = critVal(alpha = 0.122, stat = teststatmatrix, n = testn, nq = 2^9L, intervalSystem = "dyaPar", lengths = c(1, 2, 8, 16, 32), penalty = "sqrt", output = "vector"))) expect_error(stepFit(testy, alpha = 0.1, r = "s", options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 0, options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = c(100, 200), options = list(load = list()))) expect_identical(stepFit(testy, alpha = 0.1, r = 100.5, options = list(load = list())), stepFit(testy, alpha = 0.1, r = 100L, options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulat = "vector", save = list(), load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = "vector")) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vecto", save = list(), load = list()))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, intervalSystem = "dyaPar", lengths = c(2, 4, 8), penalty = "log", options = list(simulation = "vector", save = list(), load = list())), stepFit(testy, intervalSystem = "dyaPar", lengths = c(2, 4, 8), q = critVal(n = length(testy), alpha = 0.1, r = 100L, intervalSystem = "dyaPar", lengths = c(2, 4, 8), penalty = "log", options = list(simulation = "vector", save = list(), load = list())))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, lengths = 3:17, penalty = "log", options = list(simulation = "vectorIncreased", save = list(), load = list())), stepFit(testy, lengths = 3:17, q = critVal(n = length(testy), alpha = 0.1, r = 100L, lengths = 3:17, penalty = "log", options = list(simulation = "vectorIncreased", save = list(), load = list())))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, lengths = 3:17, penalty = "log", nq = 100, options = list(simulation = "vectorIncreased", save = list(), load = list())), stepFit(testy, lengths = 3:17, q = critVal(n = 100, alpha = 0.1, r = 100L, lengths = 3:17, penalty = "log", options = list(simulation = "vector", save = list(), load = list())))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, lengths = 10:13, penalty = "weights", weights = rep(1 / 4, 4), options = list(simulation = "matrix", save = list(), load = list())), stepFit(testy, lengths = 10:13, q = critVal(n = length(testy), alpha = 0.1, r = 100L, lengths = 10:13, penalty = "weights", weights = rep(1 / 4, 4), options = list(simulation = "matrix", save = list(), load = list())))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrixIncreased", save = list(), load = list())), stepFit(testy, q = critVal(n = length(testy), alpha = 0.1, r = 100L, options = list(simulation = "matrixIncreased", save = list(), load = list())))) testfile <- c(tempfile(pattern = "file", tmpdir = tempdir(), fileext = ".RDS"), tempfile(pattern = "file", tmpdir = tempdir(), fileext = ".RDS")) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(RDSfile = testfile, test = 1), load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(RDSfile = testfile), load = list(test = "test")))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(RDSfile = c(testfile, testfile)), load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(RDSfile = 10), load = list()))) teststat <- monteCarloSimulation(36L, r = 100L) teststatvec <- monteCarloSimulation(36L, r = 100L, output = "maximum") expect_identical(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(RDSfile = testfile), load = list())), stepFit(testy, alpha = 0.1, stat = teststat)) expect_identical(readRDS(testfile[1]), teststatvec) expect_identical(readRDS(testfile[2]), teststat) expect_error(critVal(100L, alpha = 0.1, r = 100L, options = list(simulation = "vector", load = list(RDSfile = 10), save = list()))) expect_error(critVal(100L, alpha = 0.1, r = 100L, options = list(simulation = "vector", load = list(RDSfile = testfile), save = list()))) expect_identical(stepFit(testy, alpha = 0.1, r = 50L, options = list(simulation = "matrix", load = list(RDSfile = testfile[1]), save = list())), stepFit(testy, alpha = 0.1, stat = teststat)) expect_identical(stepFit(testy, alpha = 0.1, r = 50L, options = list(simulation = "matrix", load = list(RDSfile = testfile[1]), save = list())), stepFit(testy, alpha = 0.1, stat = teststat)) unlink(testfile) testvariable <- c("testsavevector", "testsavematrix") testStepR <- new.env() expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(variable = c(testvariable, testvariable)), load = list(), envir = testStepR))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(variable = 10), load = list(), envir = testStepR))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(variable = testvariable), load = list(), envir = "testStepR"))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(variable = testvariable), load = list(), envir = c(testStepR, testStepR)))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(variable = testvariable), load = list(), envir = c(testStepR, 10)))) teststat <- monteCarloSimulation(36L, r = 100L, output = "maximum") expect_identical(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "vector", save = list(variable = testvariable), load = list(), envir = testStepR)), stepFit(testy, alpha = 0.1, stat = teststat)) expect_identical(get("testsavevector", envir = testStepR), teststat) expect_false(exists("testsavematrix", envir = testStepR)) remove(testsavevector, envir = testStepR) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(workspace = "matri"), load = list(), envir = testStepR))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(workspace = c("vector", "matri")), load = list(), envir = testStepR))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", load = list(workspace = "matri"), save = list(), envir = testStepR))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", load = list(workspace = c("vector", "matri")), save = list(), envir = testStepR))) teststat <- monteCarloSimulation(36L, r = 100L) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(workspace = "matrix"), load = list(), envir = testStepR)), stepFit(testy, alpha = 0.1, stat = teststat, options = list())) expect_identical(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat[[1]], teststat) expect_identical(length(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat), 1L) expect_identical(stepFit(testy, alpha = 0.1, r = 50L, options = list(simulation = "matrix", save = list(), load = list(workspace = "matrix"), envir = testStepR)), stepFit(testy, alpha = 0.1, stat = teststat, options = list())) remove(critValStepRTab, envir = testStepR) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(fileSystem = "matri"), load = list(), dirs = "testStepR"))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(fileSystem = c("vector", "matri")), load = list(), dirs = "testStepR"))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", load = list(fileSystem = "matri"), save = list(), dirs = "testStepR"))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", load = list(fileSystem = c("vector", "matri")), save = list(), dirs = "testStepR"))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(fileSystem = "matrix"), load = list(), dirs = c("testStepR", "test")))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(fileSystem = "matrix"), load = list(), dirs = 10))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, options = list(simulation = "matrix", save = list(fileSystem = "matrix"), load = list(), dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat, options = list())) expect_identical(R.cache::loadCache(attr(teststat, "keyList"), dirs = "testStepR"), teststat) expect_identical(length(list.files(file.path(R.cache::getCacheRootPath(), "testStepR"))), 1L) expect_identical(critVal(36L, alpha = 0.1, r = 50L, options = list(simulation = "matrix", save = list(), load = list(fileSystem = "matrix"), dirs = "testStepR")), critVal(36L, alpha = 0.1, stat = teststat, options = list())) expect_identical(stepFit(testy, alpha = 0.1, r = 50L, options = list(simulation = "matrix", save = list(), load = list(fileSystem = "matrix"), dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat, options = list())) unlink(file.path(R.cache::getCacheRootPath(), "testStepR"), recursive = TRUE) teststat1 <- monteCarloSimulation(36L, r = 100L, output = "maximum", lengths = c(23:34)) expect_identical(stepFit(testy, alpha = 0.1, r = 100L,lengths = c(23:34), options = list(simulation = "vector", save = list(workspace = c("matrixIncreased", "vector", "matrix"), fileSystem = c("matrixIncreased", "vector", "matrix"), RDSfile = testfile, variable = testvariable), load = list(), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat1, lengths = c(23:34), options = list())) expect_identical(readRDS(testfile[1]), teststat1) expect_false(file.exists(testfile[2])) expect_identical(get("testsavevector", envir = testStepR), teststat1) expect_false(exists("testsavematrix", envir = testStepR)) teststat2 <- monteCarloSimulation(36L, r = 100L, intervalSystem = "dyaPar") teststat3 <- monteCarloSimulation(36L, r = 100L, output = "maximum", intervalSystem = "dyaPar", lengths = 2^(1:3)) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, intervalSystem = "dyaPar", lengths = 2^(1:3), options = list(simulation = "matrix", save = list(workspace = c("vectorIncreased", "vector"), fileSystem = c("matrixIncreased", "matrix"), RDSfile = testfile, variable = testvariable), load = list(), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat2, intervalSystem = "dyaPar", lengths = 2^(1:3), options = list())) expect_identical(readRDS(testfile[1]), teststat3) expect_identical(readRDS(testfile[2]), teststat2) expect_identical(get("testsavevector", envir = testStepR), teststat3) expect_identical(get("testsavematrix", envir = testStepR), teststat2) unlink(testfile) remove(testsavevector, envir = testStepR) remove(testsavematrix, envir = testStepR) teststat <- monteCarloSimulation(63L, r = 100L) teststat4 <- monteCarloSimulation(63L, r = 100L, output = "maximum", lengths = 1:36, penalty = "log") expect_identical(stepFit(testy, alpha = 0.1, r = 100L, stat = teststat, penalty = "log", options = list(simulation = "matrixIncreased", save = list(workspace = c("matrix", "vectorIncreased"), fileSystem = c("vector", "matrix")), load = list(workspace = c("vectorIncreased", "matrix")), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat, penalty = "log", options = list())) teststat1new <- monteCarloSimulation(36L, r = 200L, output = "maximum", lengths = c(23:34)) expect_identical(critVal(36L, alpha = 0.1, r = 200L, output = "value", lengths = c(23:34), options = list(simulation = "vector", save = list(workspace = c("matrixIncreased", "vector", "matrix")), load = list(workspace = c("vector", "matrix")), envir = testStepR, dirs = "testStepR")), critVal(36L, alpha = 0.1, stat = teststat1new, output = "value", lengths = c(23:34), options = list())) expect_identical(stepFit(testy, alpha = 0.1, r = 50L, penalty = "log", options = list(simulation = "matrix", save = list(fileSystem = c("matrix", "vectorIncreased")), load = list(workspace = c("vectorIncreased", "matrix")), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat4, penalty = "log", options = list())) teststat5 <- monteCarloSimulation(36L, r = 100L, output = "maximum", intervalSystem = "dyaLen", lengths = 16L, penalty = "log") expect_identical(stepFit(testy, alpha = 0.1, r = 100L, intervalSystem = "dyaLen", lengths = 16L, penalty = "log", stat = teststat5, options = list(simulation = "vector", save = list(fileSystem = c("matrix", "vector"), workspace = c("matrix", "matrixIncreased")), load = list(workspace = c("vectorIncreased", "matrix")), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat5, intervalSystem = "dyaLen", lengths = 16L, penalty = "log", options = list())) teststat6 <- monteCarloSimulation(125L, r = 100L, intervalSystem = "dyaLen") expect_identical(stepFit(testy, alpha = 0.1, r = 100L, penalty = "weights", nq = 125L, weights = 1:6 / sum(1:6), intervalSystem = "dyaLen", options = list(simulation = "matrixIncreased", save = list(workspace = c("matrixIncreased", "vector"), fileSystem = c("matrix", "vectorIncreased")), load = list(workspace = c("vectorIncreased", "matrix")), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat6, penalty = "weights", weights = 1:6 / sum(1:6), intervalSystem = "dyaLen", options = list())) teststat6b <- monteCarloSimulation(125L, r = 100L, output = "maximum", penalty = "log", intervalSystem = "dyaLen", lengths = 2^(2:4)) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, penalty = "log", intervalSystem = "dyaLen", lengths = 2^(2:4), nq = 125L, options = list(simulation = "matrix", save = list(workspace = c("matrixIncreased", "vector"), fileSystem = c("matrix", "vectorIncreased")), load = list(workspace = c("matrixIncreased", "matrix")), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat6, penalty = "log", intervalSystem = "dyaLen", lengths = 2^(2:4), options = list())) teststat <- monteCarloSimulation(36L, r = 200L, output = "maximum", rand.gen = function(data) {rnorm(36)}) expect_identical(stepFit(testy, alpha = 0.1, r = 200L, rand.gen = function(data) {rnorm(36)}, options = list(simulation = "vector", save = list(workspace = c("matrix", "vectorIncreased", "vector")), load = list(fileSystem = c("vectorIncreased", "matrix")), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat, options = list())) teststat7 <- monteCarloSimulation(125L, r = 100L, output = "maximum", intervalSystem = "dyaLen", lengths = 2^(3:5), penalty = "log") expect_identical(critVal(120L, alpha = 0.1, r = 100L, output = "value", intervalSystem = "dyaLen", lengths = 2^(3:5), penalty = "log", nq = 125L, options = list(simulation = "vectorIncreased", save = list(workspace = c("matrixIncreased", "vectorIncreased")), load = list(workspace = c("matrix", "vector"), fileSystem = c("vectorIncreased", "matrixIncreased")), envir = testStepR, dirs = "testStepR")), critVal(120L, alpha = 0.1, stat = teststat7, output = "value", options = list(), intervalSystem = "dyaLen", lengths = 2^(3:5), penalty = "log")) expect_identical(length(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat), 5L) expect_identical(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat[[1]], teststat1new) expect_identical(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat[[2]], teststat3) expect_identical(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat[[3]], teststat4) expect_identical(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat[[4]], teststat6) expect_identical(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat[[5]], teststat7) remove(critValStepRTab, envir = testStepR) expect_identical(length(list.files(file.path(R.cache::getCacheRootPath(), "testStepR"))), 5L) expect_identical(R.cache::loadCache(attr(teststat1, "keyList"), dirs = "testStepR"), teststat1) expect_identical(R.cache::loadCache(attr(teststat2, "keyList"), dirs = "testStepR"), teststat2) expect_identical(R.cache::loadCache(attr(teststat4, "keyList"), dirs = "testStepR"), teststat4) expect_identical(R.cache::loadCache(attr(teststat5, "keyList"), dirs = "testStepR"), teststat5) expect_identical(R.cache::loadCache(attr(teststat6b, "keyList"), dirs = "testStepR"), teststat6b) unlink(file.path(R.cache::getCacheRootPath(), "testStepR"), recursive = TRUE) pathStepRdata <- NULL try(pathStepRdata <- find.package("stepRdata"), silent = TRUE) if (!is.null(pathStepRdata)) { teststat <- monteCarloSimulation(63L, intervalSystem = "dyaLen") expect_equal(stepFit(testy, alpha = 0.1, r = 100L, intervalSystem = "dyaLen", options = list(simulation = "vector", save = list(), load = list(package = TRUE), envir = testStepR, dirs = "testStepR")), stepFit(testy, alpha = 0.1, stat = teststat, intervalSystem = "dyaLen", options = list())) } expect_error(stepFit(testy, alpha = 0.1, test = 1, options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, sd = "s", options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, sd = 0, options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, sd = c(1, 2), options = list(load = list()))) expect_error(supressWarning(stepFit(testy, alpha = 0.1, r = 100L, seed = "s", options = list(load = list())))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, seed = c(1, 2), options = list(load = list())), stepFit(testy, alpha = 0.1, r = 100L, seed = 1L, options = list(load = list()))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, seed = 100.5, options = list(load = list())), stepFit(testy, alpha = 0.1, r = 100L, seed = 100L, options = list(load = list()))) teststat <- monteCarloSimulation(n = 63L, r = 100L) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, seed = 63L, options = list(load = list())), stepFit(testy, alpha = 0.1, r = 100L, stat = teststat)) teststat <- monteCarloSimulation(n = 36L, r = 100L) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, seed = 36L, options = list(load = list(), simulation = "matrix")), stepFit(testy, alpha = 0.1, r = 100L, stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, r = 100L, rand.gen = 10, options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, rand.gen = function(data, n) {rnorm(10)}, options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, rand.gen = function(data) {rnorm(10)}, options = list(load = list()))) teststat <- monteCarloSimulation(n = 63L, r = 100L) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, rand.gen = function(data) {rnorm(63)}, options = list(load = list())), stepFit(testy, alpha = 0.1, r = 100L, stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, r = 100L, messages = "s", options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, messages = c(10, 20), options = list(load = list()))) expect_error(stepFit(testy, alpha = 0.1, r = 100L, messages = 0, options = list(load = list()))) expect_identical(suppressMessages(stepFit(testy, alpha = 0.1, r = 100L, messages = 10.5, options = list(load = list()))), suppressMessages(stepFit(testy, alpha = 0.1, r = 100L, messages = 10L, options = list(load = list())))) }) test_that("arguments confband and jumpint work and are tested", { testn <- 40L testy <- c(rnorm(10L), rnorm(10L, 5), rnorm(10L, 10), rnorm(10L, 15)) testq <- rep(4, testn) expect_error(stepFit(y = testy, q = testq, confband = 1)) expect_error(stepFit(y = testy, q = testq, confband = c(FALSE, FALSE))) expect_error(stepFit(y = testy, q = testq, confband = as.logical(NA))) expect_error(stepFit(y = testy, q = testq, jumpint = "s")) expect_error(stepFit(y = testy, q = testq, jumpint = c(TRUE, TRUE))) expect_error(stepFit(y = testy, q = testq, jumpint = as.logical(NA))) expect_identical(stepFit(y = testy, q = testq), stepFit(y = testy, q = testq, confband = FALSE, jumpint = FALSE)) expect_identical(stepFit(y = testy, q = testq, confband = FALSE), stepFit(y = testy, q = testq, confband = FALSE, jumpint = FALSE)) expect_identical(stepFit(y = testy, q = testq, jumpint = FALSE), stepFit(y = testy, q = testq, confband = FALSE, jumpint = FALSE)) expect_identical(stepFit(y = testy, q = testq, confband = TRUE), stepFit(y = testy, q = testq, confband = TRUE, jumpint = TRUE)) expect_identical(stepFit(y = testy, q = testq, confband = TRUE, jumpint = FALSE), stepFit(y = testy, q = testq, confband = TRUE, jumpint = TRUE)) s <- stepFit(testy, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, sd = 1, confband = FALSE, jumpint = TRUE) s3 <- stepFit(testy, q = testq, sd = 1, confband = FALSE, jumpint = FALSE) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) testn <- 16L testy <- rnorm(testn) testq <- rep(5, testn) s <- stepFit(testy, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, sd = 1, confband = FALSE, jumpint = TRUE) s3 <- stepFit(testy, q = testq, sd = 1, confband = FALSE, jumpint = FALSE) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) testn <- 16L testy <- c(rnorm(testn / 2), rnorm(testn / 2, 20)) testq <- rep(5, testn) s <- stepFit(testy, q = testq, sd = 1, confband = TRUE) bounds <- computeBounds(testy, q = testq, sd = 1) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9, sd = 1) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9, sd = 1) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, sd = 1, confband = FALSE, jumpint = TRUE) s3 <- stepFit(testy, q = testq, sd = 1, confband = FALSE, jumpint = FALSE) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) }) test_that("... is tested and works", { testn <- 22L testy <- rnorm(testn) testq <- rep(1, testn) expect_error(stepFit(testy, q = testq, std = 1)) expect_error(stepFit(testy, sd = "s", q = testq)) expect_error(stepFit(testy, sd = c(1, 2), q = testq)) expect_error(stepFit(testy, sd = NA, q = testq)) expect_error(stepFit(c(1, 2), q = c(2, 2))) expect_error(stepFit(testy, sd = Inf, q = testq)) expect_error(stepFit(testy, sd = 0, q = testq)) expect_error(stepFit(testy, sd = -0.1, q = testq)) expect_identical(stepFit(testy, q = testq), stepFit(testy, q = testq, sd = sdrobnorm(testy))) expect_identical(stepFit(testy, q = testq, sd = 1L), stepFit(testy, q = testq, sd = 1)) }) test_that("family 'hsmuce' works", { testn <- 40L testy <- rnorm(testn) testq <- rep(4, 5) expect_error(stepFit(family = "hsmuce", q = testq)) expect_identical(stepFit(testy, family = "hsmuce", q = testq), stepFit(testy, x = 1:testn, x0 = 0, family = "hsmuce", intervalSystem = "dyaPar", lengths = 2^(1:5), q = testq, confband = FALSE, jumpint = FALSE)) expect_error(stepFit(as.integer(testy), family = "hsmuce", q = testq)) expect_error(stepFit(c(testy, "s"), family = "hsmuce", q = testq)) expect_error(stepFit(c(rnorm(10), NA), family = "hsmuce", q = testq)) s <- stepFit(testy, q = testq, family = "hsmuce", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce") testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testy <- c(rnorm(40, 0), rnorm(5, 10), rnorm(5, 20), rnorm(5, 30), rnorm(40, 40)) testn <- length(testy) testq <- c(1e4, 1e2, rep(3, 4)) testx <- 1:testn / 1e4 s <- stepFit(testy, x = testx, q = testq, family = "hsmuce", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce") testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testy <- c(rnorm(20, 0), rnorm(5, 10), rnorm(5, 20), rnorm(5, 30), rnorm(20, 40)) testn <- length(testy) testq <- c(1e4, 1e2, rep(3, 3)) testx <- 1:testn / 5.34 s <- stepFit(testy, x = testx, x0 = -1, q = testq, family = "hsmuce", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce") testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testn <- 23L testy <- c(rnorm(10, 1, 0.23), rnorm(testn - 10, -1, 0.34)) testq <- c(1e5, 1e4, 1e3, 1e2, 1e1, rep(5, 18)) expect_identical(stepFit(testy, family = "hsmuce", q = testq), stepFit(testy, family = "hsmuce", q = testq, intervalSystem = "dyaPar")) testx <- 1:testn s <- stepFit(testy, q = testq, family = "hsmuce", intervalSystem = "all", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce", intervalSystem = "all") testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testx <- 1:testn s <- stepFit(testy, q = testq, family = "hsmuce", intervalSystem = "dyaLen", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce", intervalSystem = "dyaLen") testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testn <- 36L testy <- c(rnorm(12, 1, 0.13), rnorm(12, -1, 0.23), rnorm(12, 1, 0.43)) testq <- 2 expect_error(stepFit(testy, family = "hsmuce", lengths = 1L, q = testq)) expect_error(stepFit(testy, family = "hsmuce", lengths = 64L, q = testq)) expect_error(stepFit(testy, family = "hsmuce", intervalSystem = "dyaLen", lengths = 1L, q = testq)) expect_error(stepFit(testy, family = "hsmuce", intervalSystem = "dyaLen", lengths = 64L, q = testq)) expect_error(stepFit(testy, family = "hsmuce", intervalSystem = "all", lengths = 1L, q = testq)) expect_error(stepFit(testy, family = "hsmuce", intervalSystem = "all", lengths = 64L, q = testq)) testq <- c(1e5, 1e4, 1e3, 1e2, 1e1, rep(4, 31)) expect_equal(stepFit(testy, family = "hsmuce", intervalSystem = "all", lengths = 2^(1:5), q = testq), stepFit(testy, family = "hsmuce", intervalSystem = "dyaLen", q = testq)) expect_error(stepFit(testy, family = "hsmuce", q = testq, sd = 1)) testq <- c(1e5, 8:1) testlengths = c(2, 5:10, 35:36) testx <- 1:testn s <- stepFit(testy, q = testq, family = "hsmuce", intervalSystem = "all", lengths = testlengths, confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce", intervalSystem = "all", lengths = testlengths) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testq <- c(1e5, 1e2, 5) testlengths = c(2, 4, 16) testx <- 1:testn s <- stepFit(testy, q = testq, family = "hsmuce", intervalSystem = "dyaLen", lengths = testlengths, confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce", intervalSystem = "dyaLen", lengths = testlengths) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testq <- c(1e5, 4) testlengths = c(2, 8) testx <- 1:testn s <- stepFit(testy, q = testq, family = "hsmuce", lengths = testlengths, confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce", lengths = testlengths) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument q is tested and works for family 'hsmuce'", { testn <- 36L testy <- c(rnorm(12, 1, 0.13), rnorm(12, -1, 0.23), rnorm(12, 1, 0.43)) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "hsmuce") expect_warning(ret <- stepFit(testy, family = "hsmuce")) expect_identical(ret, stepFit(testy, family = "hsmuce", alpha = 0.5)) expect_error(stepFit(testy, q = Inf, alpha = 0.1, stat = teststat, family = "hsmuce")) expect_error(stepFit(testy, q = rep(1, 6), alpha = 0.1, stat = teststat, family = "hsmuce")) expect_error(stepFit(testy, q = rep(1, 4), alpha = 0.1, stat = teststat, family = "hsmuce")) testq <- 1:5 attr(testq, "n") <- "s" expect_error(stepFit(testy, q = testq, alpha = 0.1, stat = teststat, family = "hsmuce")) attr(testq, "n") <- 35L expect_error(stepFit(testy, q = testq, alpha = 0.1, stat = teststat, family = "hsmuce")) expect_identical(stepFit(testy, q = 3, family = "hsmuce", intervalSystem = "all", lengths = c(2:3, 8:23), penalty = "sqrt"), stepFit(testy, family = "hsmuce", intervalSystem = "all", lengths = c(2:3, 8:23), q = critVal(q = 3, family = "hsmuce", intervalSystem = "all", lengths = c(2:3, 8:23), n = testn, penalty = "sqrt"))) expect_identical(stepFit(testy, q = 3, intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), nq = 45L, penalty = "none"), stepFit(testy, intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), q = critVal(q = 3, intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), n = testn, nq = 45L, penalty = "none"))) expect_identical(stepFit(testy, q = 3, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), penalty = "log"), stepFit(testy, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), q = critVal(q = 3, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), n = testn, penalty = "log"))) testq <- 1:45 attr(testq, "n") <- 45L expect_identical(stepFit(testy, family = "hsmuce", intervalSystem = "all", q = testq), stepFit(testy, family = "hsmuce", intervalSystem = "all", q = 2:36)) testq <- 2:45 attr(testq, "n") <- 45L expect_identical(stepFit(testy, family = "hsmuce", intervalSystem = "all", q = testq), stepFit(testy, family = "hsmuce", intervalSystem = "all", q = 2:36)) testq <- 1:36 attr(testq, "n") <- 45L expect_identical(stepFit(testy, family = "hsmuce", intervalSystem = "all", q = testq), stepFit(testy, family = "hsmuce", intervalSystem = "all", q = 2:36)) expect_identical(stepFit(testy, family = "hsmuce", intervalSystem = "all", q = 1:36, lengths = 3:23), stepFit(testy, family = "hsmuce", intervalSystem = "all", q = 3:23, lengths = 3:23)) expect_identical(stepFit(testy, q = 1:36, family = "hsmuce", intervalSystem = "dyaLen"), stepFit(testy, q = 2^(1:5), family = "hsmuce", intervalSystem = "dyaLen")) testq <- 2^(1:6) attr(testq, "n") <- 64L expect_identical(stepFit(testy, q = testq, family = "hsmuce", intervalSystem = "dyaPar"), stepFit(testy, q = 2^(1:5), family = "hsmuce", intervalSystem = "dyaPar")) ret <- stepFit(testy, family = "hsmuce", alpha = 0.1, stat = teststat) expect_identical(ret, stepFit(testy, family = "hsmuce", alpha = 0.1, stat = teststat, nq = testn, intervalSystem = "dyaPar", lengths = 2^(1:5), penalty = "weights", weights = rep(0.2, 5))) expect_identical(ret, stepFit(testy, family = "hsmuce", q = critVal(family = "hsmuce", alpha = 0.1, stat = teststat, n = testn))) expect_error(stepFit(testy, family = "hsmuce", alpha = "s", stat = teststat)) expect_identical(stepFit(testy, family = "hsmuce", alpha = 0.075, stat = teststat), stepFit(testy, family = "hsmuce", q = critVal(alpha = 0.075, family = "hsmuce", stat = teststat, n = testn))) expect_error(stepFit(testy, family = "hsmuce", alpha = 0.1, stat = teststat, n = testn)) expect_error(stepFit(testy, alpha = 0.1, family = "hsmuce", weights = NA, stat = teststat)) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "hsmuce", intervalSystem = "all") expect_identical(stepFit(testy, family = "hsmuce", alpha = 0.14, penalty = "log", stat = teststat, intervalSystem = "all", lengths = c(2, 5, 8, 23)), stepFit(testy, family = "hsmuce", intervalSystem = "all", lengths = c(2, 5, 8, 23), q = critVal(alpha = 0.14, family = "hsmuce", intervalSystem = "all", lengths = c(2, 5, 8, 23), penalty = "log", stat = teststat, n = testn))) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "hsmuce", intervalSystem = "dyaLen") expect_identical(stepFit(testy, family = "hsmuce", alpha = 0.034, penalty = "none", stat = teststat, intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32)), stepFit(testy, family = "hsmuce", intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), q = critVal(alpha = 0.034, family = "hsmuce", intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), penalty = "none", stat = teststat, n = testn))) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "hsmuce") expect_identical(stepFit(testy, family = "hsmuce", alpha = 0.56, penalty = "weights", weights = c(0.2, 0.3, 0.4, 0.1), stat = teststat, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32)), stepFit(testy, family = "hsmuce", intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), q = critVal(alpha = 0.56, family = "hsmuce", intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), penalty = "weights", weights = c(0.2, 0.3, 0.4, 0.1), stat = teststat, n = testn))) expect_identical(stepFit(testy, family = "hsmuce", alpha = 0.014, stat = teststat, nq = 100L, intervalSystem = "dyaPar", lengths = 2^c(1:3, 5)), stepFit(testy, family = "hsmuce", intervalSystem = "dyaPar", lengths = 2^c(1:3, 5), q = critVal(alpha = 0.014, family = "hsmuce", intervalSystem = "dyaPar", lengths = 2^c(1:3, 5), stat = teststat, n = testn, nq = 100L))) expect_error(stepFit(testy, family = "hsmuce", alpha = 0.1, output = "vector", stat = teststat)) expect_error(stepFit(testy, family = "hsmuce", alpha = 0.1, data = 1, stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, family = "hsmuce", stat = rnorm(1e4))) teststatmatrix <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "all", family = "hsmuce") teststatvector <- monteCarloSimulation(n = 36L, r = 100L, output = "maximum", intervalSystem = "all", penalty = "sqrt", family = "hsmuce") expect_identical(stepFit(testy, family = "hsmuce", intervalSystem = "all", penalty = "sqrt", alpha = 0.05, stat = teststatvector), stepFit(testy, family = "hsmuce", intervalSystem = "all", q = critVal(family = "hsmuce", intervalSystem = "all", alpha = 0.05, penalty = "sqrt", stat = teststatmatrix, n = testn))) teststatmatrix <- monteCarloSimulation(n = 36L, r = 100L, intervalSystem = "dyaPar", family = "hsmuce") teststatvector <- monteCarloSimulation(n = 36L, r = 100L, output = "maximum", intervalSystem = "dyaPar", penalty = "sqrt", family = "hsmuce", lengths = c(2, 8, 16, 32)) expect_identical(stepFit(testy, alpha = 0.122, stat = teststatvector, nq = 2^9L, family = "hsmuce", intervalSystem = "dyaPar", lengths = c(2, 8, 16, 32), penalty = "sqrt"), stepFit(testy, family = "hsmuce", intervalSystem = "dyaPar", lengths = c(2, 8, 16, 32), q = critVal(alpha = 0.122, stat = teststatmatrix, n = testn, nq = 2^9L, family = "hsmuce", intervalSystem = "dyaPar", lengths = c(2, 8, 16, 32), penalty = "sqrt"))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, intervalSystem = "dyaPar", lengths = c(2, 4, 8), penalty = "log", family = "hsmuce", options = list(simulation = "vector", save = list(), load = list())), stepFit(testy, intervalSystem = "dyaPar", family = "hsmuce", lengths = c(2, 4, 8), q = critVal(n = length(testy), alpha = 0.1, r = 100L, intervalSystem = "dyaPar", lengths = c(2, 4, 8), penalty = "log", family = "hsmuce", options = list(simulation = "vector", save = list(), load = list())))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, family = "hsmuce", weights = c(0.4, 0.2, 0.1, 0.1, 0.2), options = list(simulation = "matrixIncreased", save = list(), load = list())), stepFit(testy, family = "hsmuce", q = critVal(n = length(testy), alpha = 0.1, r = 100L, family = "hsmuce" , weights = c(0.4, 0.2, 0.1, 0.1, 0.2), options = list(simulation = "matrixIncreased", save = list(), load = list())))) }) test_that("arguments confband and jumpint work and are tested for family 'hsmuce'", { testn <- 40L testy <- c(rnorm(10L, 0.5), rnorm(10L, 5, 0.3), rnorm(10L, 10, 0.2), rnorm(10L, 15, 0.6)) testq <- c(1e5, 1e4, 1e3, 1e2, 1e1, rep(5, 35)) s <- stepFit(testy, q = testq, family = "hsmuce", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce") testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, family = "hsmuce", confband = FALSE, jumpint = TRUE) s3 <- stepFit(testy, q = testq, family = "hsmuce", confband = FALSE, jumpint = FALSE) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) testn <- 16L testy <- rnorm(testn) testq <- c(1e5, 1e4, 1e3, 1e2, 1e1, rep(5, 11)) s <- stepFit(testy, q = testq, family = "hsmuce", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce") testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, family = "hsmuce", confband = FALSE, jumpint = TRUE) s3 <- stepFit(testy, q = testq, family = "hsmuce", confband = FALSE, jumpint = FALSE) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) testn <- 16L testy <- c(rnorm(testn / 2), rnorm(testn / 2, 20, 0.5)) testq <- c(1e5, 1e4, 1e3, 1e2, 1e1, rep(5, 11)) s <- stepFit(testy, q = testq, family = "hsmuce", confband = TRUE) bounds <- computeBounds(testy, q = testq, family = "hsmuce") testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, family = "hsmuce", confband = FALSE, jumpint = TRUE) s3 <- stepFit(testy, q = testq, family = "hsmuce", confband = FALSE, jumpint = FALSE) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) }) test_that("family 'mDependentPS' works", { testn <- 40L testy <- rnorm(testn) testq <- rep(4, testn) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.8, 0.5, 0.3), lag.max = 3)) expect_error(stepFit(family = "mDependentPS", q = testq, covariances = testcovariances)) expect_identical(stepFit(testy, family = "mDependentPS", q = testq, covariances = testcovariances), stepFit(testy, x = 1:testn, x0 = 0, family = "mDependentPS", intervalSystem = "dyaLen", lengths = 2^(0:5), q = testq, confband = FALSE, jumpint = FALSE, covariances = testcovariances)) expect_error(stepFit(as.integer(testy), family = "mDependentPS", q = testq, covariances = testcovariances)) expect_error(stepFit(c(testy, "s"), family = "mDependentPS", q = testq, covariances = testcovariances)) expect_error(stepFit(c(rnorm(10), NA), family = "mDependentPS", q = testq, covariances = testcovariances)) s <- stepFit(testy, q = testq, family = "mDependentPS", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", covariances = testcovariances) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testy <- c(rnorm(40, 0), rnorm(5, 10), rnorm(5, 20), rnorm(5, 30), rnorm(40, 40)) testn <- length(testy) testq <- rep(4, testn) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.8, 0.4, 0.2), lag.max = 3)) testx <- 1:testn / 1e4 s <- stepFit(testy, x = testx, q = testq, family = "mDependentPS", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", covariances = testcovariances) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testy <- c(rnorm(20, 0), rnorm(5, 10), rnorm(5, 20), rnorm(5, 30), rnorm(20, 40)) testn <- length(testy) testq <- rep(5, 55) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.9, 0.7, 0.5, 0.3), lag.max = 4)) testx <- 1:testn / 5.34 s <- stepFit(testy, x = testx, x0 = -1, q = testq, family = "mDependentPS", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", covariances = testcovariances) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testn <- 23L testy <- c(rnorm(10, 1, 0.23), rnorm(testn - 10, -1, 0.34)) testq <- rep(4, 23) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.5, 0.3), lag.max = 2)) expect_identical(stepFit(testy, family = "mDependentPS", q = testq, covariances = testcovariances), stepFit(testy, family = "mDependentPS", q = testq, intervalSystem = "dyaLen", covariances = testcovariances)) testx <- 1:testn s <- stepFit(testy, q = testq, family = "mDependentPS", intervalSystem = "all", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", intervalSystem = "all", covariances = testcovariances) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testx <- 1:testn s <- stepFit(testy, q = testq, family = "mDependentPS", intervalSystem = "dyaPar", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", intervalSystem = "dyaPar", covariances = testcovariances) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testn <- 36L testy <- c(rnorm(12, 1, 0.13), rnorm(12, -1, 0.23), rnorm(12, 1, 0.43)) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.9, 0.65, 0.34, 0.27, 0.1, 0.05), lag.max = 6)) / 10 testq <- rep(4, 36) expect_equal(stepFit(testy, family = "mDependentPS", intervalSystem = "all", lengths = 2^(0:5), q = testq, covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "dyaLen", q = testq, covariances = testcovariances)) testq <- 9:1 testlengths = c(1, 5:10, 35:36) testx <- 1:testn s <- stepFit(testy, q = testq, family = "mDependentPS", intervalSystem = "all", lengths = testlengths, confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", intervalSystem = "all", lengths = testlengths, covariances = testcovariances) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testq <- rep(4, 4) testlengths = c(1, 2, 4, 16) testx <- 1:testn s <- stepFit(testy, q = testq, family = "mDependentPS", lengths = testlengths, confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", lengths = testlengths, covariances = testcovariances) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) testq <- c(4, 4) testlengths = c(1, 8) testx <- 1:testn s <- stepFit(testy, q = testq, family = "mDependentPS", intervalSystem = "dyaPar", lengths = testlengths, confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", intervalSystem = "dyaPar", lengths = testlengths, covariances = testcovariances) testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("argument q is tested and works for family 'mDependentPS'", { testn <- 36L testy <- c(rnorm(12, 10, 0.13), rnorm(12, -10, 0.13), rnorm(12, 10, 0.13)) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.8, 0.5, 0.3), lag.max = 3)) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances) expect_warning(ret <- stepFit(testy, family = "mDependentPS", covariances = testcovariances)) expect_identical(ret, stepFit(testy, family = "mDependentPS", covariances = testcovariances, alpha = 0.5)) expect_error(stepFit(testy, q = rep(1, 7), alpha = 0.1, stat = teststat, family = "mDependentPS", covariances = testcovariances)) expect_error(stepFit(testy, q = rep(1, 4), alpha = 0.1, stat = teststat, family = "mDependentPS", covariances = testcovariances)) expect_identical(stepFit(testy, q = 3, family = "mDependentPS", intervalSystem = "all", lengths = c(2:3, 8:23), penalty = "sqrt", covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "all", lengths = c(2:3, 8:23), covariances = testcovariances, q = critVal(q = 3, family = "mDependentPS", intervalSystem = "all", lengths = c(2:3, 8:23), n = testn, penalty = "sqrt", covariances = testcovariances))) expect_identical(stepFit(testy, q = 3, family = "mDependentPS", intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), nq = 45L, penalty = "none", covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), covariances = testcovariances, q = critVal(q = 3, family = "mDependentPS", intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), n = testn, nq = 45L, penalty = "none", covariances = testcovariances))) expect_identical(stepFit(testy, q = 3, family = "mDependentPS", intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), penalty = "log", covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), covariances = testcovariances, q = critVal(q = 3, family = "mDependentPS", intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), n = testn, penalty = "log", covariances = testcovariances))) testq <- 1:45 attr(testq, "n") <- 45L expect_identical(stepFit(testy, family = "mDependentPS", intervalSystem = "all", q = testq, covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "all", q = 1:36, covariances = testcovariances)) expect_identical(stepFit(testy, family = "mDependentPS", intervalSystem = "all", q = 1:36, lengths = 3:23, covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "all", q = 3:23, lengths = 3:23, covariances = testcovariances)) expect_identical(stepFit(testy, q = 1:36, family = "mDependentPS", intervalSystem = "dyaLen", covariances = testcovariances), stepFit(testy, q = 2^(0:5), family = "mDependentPS", intervalSystem = "dyaLen", covariances = testcovariances)) testq <- 2^(0:6) attr(testq, "n") <- 64L expect_identical(stepFit(testy, q = testq, family = "mDependentPS", intervalSystem = "dyaPar", covariances = testcovariances), stepFit(testy, q = 2^(0:5), family = "mDependentPS", intervalSystem = "dyaPar", covariances = testcovariances)) ret <- stepFit(testy, family = "mDependentPS", alpha = 0.1, stat = teststat, covariances = testcovariances) expect_identical(ret, stepFit(testy, family = "mDependentPS", alpha = 0.1, stat = teststat, nq = testn, intervalSystem = "dyaLen", lengths = 2^(0:5), penalty = "sqrt", covariances = testcovariances)) expect_identical(ret, stepFit(testy, family = "mDependentPS", covariances = testcovariances, q = critVal(family = "mDependentPS", alpha = 0.1, stat = teststat, n = testn, covariances = testcovariances))) expect_error(stepFit(testy, family = "mDependentPS", alpha = "s", stat = teststat, covariances = testcovariances)) expect_identical(stepFit(testy, family = "mDependentPS", alpha = 0.075, stat = teststat, covariances = testcovariances), stepFit(testy, family = "mDependentPS", covariances = testcovariances, q = critVal(alpha = 0.075, family = "mDependentPS", stat = teststat, n = testn, covariances = testcovariances))) expect_error(stepFit(testy, family = "mDependentPS", alpha = 0.1, stat = teststat, n = testn, covariances = testcovariances)) expect_error(stepFit(testy, alpha = 0.1, family = "mDependentPS", penalty = "weights", weights = NA, stat = teststat, covariances = testcovariances)) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances, intervalSystem = "all") expect_identical(stepFit(testy, family = "mDependentPS", alpha = 0.14, penalty = "log", stat = teststat, intervalSystem = "all", lengths = c(2, 5, 8, 23), covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "all", lengths = c(2, 5, 8, 23), covariances = testcovariances, q = critVal(alpha = 0.14, family = "mDependentPS", intervalSystem = "all", lengths = c(2, 5, 8, 23), penalty = "log", stat = teststat, n = testn, covariances = testcovariances))) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances, intervalSystem = "dyaLen") expect_identical(stepFit(testy, family = "mDependentPS", alpha = 0.034, penalty = "none", stat = teststat, intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), covariances = testcovariances, q = critVal(alpha = 0.034, family = "mDependentPS", intervalSystem = "dyaLen", lengths = c(2, 4, 8, 32), penalty = "none", stat = teststat, n = testn, covariances = testcovariances))) teststat <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances, intervalSystem = "dyaPar") expect_identical(stepFit(testy, family = "mDependentPS", alpha = 0.56, penalty = "weights", weights = c(0.2, 0.3, 0.4, 0.1), stat = teststat, intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), covariances = testcovariances, q = critVal(alpha = 0.56, family = "mDependentPS", intervalSystem = "dyaPar", lengths = c(2, 4, 8, 32), penalty = "weights", weights = c(0.2, 0.3, 0.4, 0.1), stat = teststat, n = testn, covariances = testcovariances))) expect_identical(stepFit(testy, family = "mDependentPS", alpha = 0.014, stat = teststat, nq = 100L, intervalSystem = "dyaPar", lengths = 2^c(0:3, 5), covariances = testcovariances), stepFit(testy, family = "mDependentPS", intervalSystem = "dyaPar", lengths = 2^c(0:3, 5), covariances = testcovariances, q = critVal(alpha = 0.014, family = "mDependentPS", intervalSystem = "dyaPar", lengths = 2^c(0:3, 5), stat = teststat, n = testn, nq = 100L, covariances = testcovariances))) expect_error(stepFit(testy, family = "mDependentPS", alpha = 0.1, output = "vector", stat = teststat, covariances = testcovariances)) expect_error(stepFit(testy, family = "mDependentPS", alpha = 0.1, data = 1, stat = teststat, covariances = testcovariances)) expect_error(stepFit(testy, family = "mDependentPS", covariances = testcovariances, alpha = 0.1, output = "vector", stat = teststat)) expect_error(stepFit(testy, family = "mDependentPS", covariances = testcovariances, alpha = 0.1, data = 1, stat = teststat)) expect_error(stepFit(testy, alpha = 0.1, family = "mDependentPS", covariances = testcovariances, stat = rnorm(1e4), penalty = "weights")) teststatmatrix <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances, intervalSystem = "all") teststatvector <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances, output = "maximum", intervalSystem = "all") expect_identical(stepFit(testy, family = "mDependentPS", covariances = testcovariances, intervalSystem = "all", penalty = "sqrt", alpha = 0.078, stat = teststatvector), stepFit(testy, family = "mDependentPS", covariances = testcovariances, intervalSystem = "all", q = critVal(family = "mDependentPS", covariances = testcovariances, intervalSystem = "all", alpha = 0.078, penalty = "sqrt", stat = teststatmatrix, n = testn))) teststatmatrix <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances) teststatvector <- monteCarloSimulation(n = 36L, r = 100L, family = "mDependentPS", covariances = testcovariances, output = "maximum", lengths = c(2, 8, 16, 32)) expect_identical(stepFit(testy, alpha = 0.122, stat = teststatvector, nq = 2^9L, family = "mDependentPS", covariances = testcovariances, lengths = c(2, 8, 16, 32)), stepFit(testy, family = "mDependentPS", covariances = testcovariances, lengths = c(2, 8, 16, 32), q = critVal(alpha = 0.122, stat = teststatmatrix, n = testn, nq = 2^9L, family = "mDependentPS", covariances = testcovariances, lengths = c(2, 8, 16, 32)))) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, lengths = 3:17, penalty = "log", nq = 100, family = "mDependentPS", covariances = testcovariances, intervalSystem = "all", options = list(simulation = "vectorIncreased", save = list(), load = list())), stepFit(testy, lengths = 3:17, family = "mDependentPS", covariances = testcovariances, intervalSystem = "all", q = critVal(n = 100, alpha = 0.1, r = 100L, lengths = 3:17, intervalSystem = "all", penalty = "log", family = "mDependentPS", covariances = testcovariances, options = list(simulation = "vectorIncreased", save = list(), load = list())))) testStepR <- new.env() teststat <- monteCarloSimulation(36L, r = 100L, family = "mDependentPS", covariances = testcovariances) expect_identical(stepFit(testy, alpha = 0.1, r = 100L, family = "mDependentPS", covariances = testcovariances, options = list(simulation = "matrix", save = list(workspace = "matrix"), load = list(), envir = testStepR)), stepFit(testy, alpha = 0.1, stat = teststat, family = "mDependentPS", covariances = testcovariances, options = list())) expect_identical(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat[[1]], teststat) expect_identical(length(get("critValStepRTab", envir = testStepR, inherits = FALSE)$stat), 1L) expect_identical(critVal(36L, alpha = 0.1, r = 50L, output = "vector", family = "mDependentPS", covariances = testcovariances + 1e-12, options = list(simulation = "matrix", save = list(), load = list(workspace = "matrix"), envir = testStepR)), critVal(36L, alpha = 0.1, stat = teststat, output = "vector", options = list(), family = "mDependentPS", covariances = testcovariances)) remove(critValStepRTab, envir = testStepR) }) test_that("arguments confband and jumpint work and are tested for family 'mDependentPS'", { testn <- 40L testy <- c(rnorm(10L, 0.5), rnorm(10L, 5, 0.3), rnorm(10L, 10, 0.2), rnorm(10L, 15, 0.6)) testq <- rep(5, 40) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.8, 0.5, 0.3), lag.max = 3)) s <- stepFit(testy, q = testq, family = "mDependentPS", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", covariances = testcovariances) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, family = "mDependentPS", confband = FALSE, jumpint = TRUE, covariances = testcovariances) s3 <- stepFit(testy, q = testq, family = "mDependentPS", confband = FALSE, jumpint = FALSE, covariances = testcovariances) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) testn <- 16L testy <- rnorm(testn) testq <- rep(5, 16) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.8, 0.7, 0.6, 0.5, 0.4, 0.3), lag.max = 6)) s <- stepFit(testy, q = testq, family = "mDependentPS", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", covariances = testcovariances) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, family = "mDependentPS", confband = FALSE, jumpint = TRUE, covariances = testcovariances) s3 <- stepFit(testy, q = testq, family = "mDependentPS", confband = FALSE, jumpint = FALSE, covariances = testcovariances) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) testn <- 16L testy <- c(rnorm(testn / 2), rnorm(testn / 2, 20, 0.5)) testq <- rep(5, 16) testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.8, 0.8, 0.8), lag.max = 3)) s <- stepFit(testy, q = testq, family = "mDependentPS", confband = TRUE, covariances = testcovariances) bounds <- computeBounds(testy, q = testq, family = "mDependentPS", covariances = testcovariances) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s2 <- stepFit(testy, q = testq, family = "mDependentPS", confband = FALSE, jumpint = TRUE, covariances = testcovariances) s3 <- stepFit(testy, q = testq, family = "mDependentPS", confband = FALSE, jumpint = FALSE, covariances = testcovariances) attr(s2, "confband") <- attr(s, "confband") expect_equal(s2, s, tolerance = 1e-12) s3$leftEndLeftBound <- s$leftEndLeftBound s3$leftEndRightBound <- s$leftEndRightBound s3$rightEndLeftBound <- s$rightEndLeftBound s3$rightEndRightBound <- s$rightEndRightBound s3$leftIndexLeftBound <- s$leftIndexLeftBound s3$leftIndexRightBound <- s$leftIndexRightBound s3$rightIndexLeftBound <- s$rightIndexLeftBound s3$rightIndexRightBound <- s$rightIndexRightBound attr(s3, "confband") <- attr(s, "confband") expect_equal(s3, s, tolerance = 1e-12) }) test_that("arguments in ... are tested and work for family mDependentPS", { testn <- 17L testcovariances <- as.numeric(ARMAacf(ar = c(), ma = c(0.8, 0.5, 0.3), lag.max = 3)) testcorrelations <- testcovariances / testcovariances[1] testsignal <- list(leftIndex = c(1L, 13L), rightIndex = c(12L, testn), value = c(0, -1)) testfilter <- list(acf = testcorrelations) class(testfilter) <- c("lowpassFilter", class(testfilter)) testy <- as.numeric(arima.sim(n = testn, list(ar = c(), ma = c(0.8, 0.5, 0.3)), sd = 1)) testq = 17:1 / 3 expect_error(stepFit(testy, family = "mDependentPS", q = testq, covariances = testcovariances, std = 1)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, correlations = testcorrelations, std = 1)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, filter = testfilter, std = 1)) expect_error(stepFit(testy, family = "mDependentPS", intervalsystem = "all")) expect_error(stepFit(testy, family = "mDependentPS", q = testq)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = 1)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, covariances = c(testcovariances, "s"))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, covariances = c(testcovariances, NA))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, covariances = c(testcovariances, Inf))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, covariances = c(0.01, testcovariances))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, covariances = c(-1, testcovariances))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, correlations = c(testcorrelations, "s"))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, correlations = c(testcorrelations, NA))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, correlations = c(testcorrelations, Inf))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, correlations = c(testcorrelations, 1.1))) expect_error(cstepFit(testy, family = "mDependentPS", q = testq, correlations = c(0.99, testcorrelations[-1]))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = "s", correlations = testcorrelations)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = c(1, 2), correlations = testcorrelations)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = NA, correlations = testcorrelations)) expect_error(stepFit(c(1, 2), q = c(5, 4), family = "mDependentPS", correlations = testcorrelations)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = Inf, correlations = testcorrelations)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = 0, correlations = testcorrelations)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = -0.1, correlations = testcorrelations)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, filter = list(param = list(acf = testcorrelations)))) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = "s", filter = testfilter)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = c(1, 2), filter = testfilter)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = NA, filter = testfilter)) expect_error(stepFit(c(1, 2), q = c(5, 4), family = "mDependentPS", filter = testfilter)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = Inf, filter = testfilter)) expect_error(stepFit(testy, family = "mDependentPS", q = testq, sd = 0, filter = testfilter)) expect_error(cstepFit(testy, family = "mDependentPS", q = testq, sd = -0.1, filter = testfilter)) expect_identical(stepFit(testy, family = "mDependentPS", q = testq, covariances = sdrobnorm(testy, lag = 4)^2 * testcorrelations), stepFit(testy, family = "mDependentPS", q = testq, correlations = testcorrelations)) expect_identical(stepFit(testy, family = "mDependentPS", q = testq, covariances = 1.1^2 * testcorrelations), stepFit(testy, family = "mDependentPS", q = testq, correlations = testcorrelations, sd = 1.1)) expect_identical(stepFit(testy, family = "mDependentPS", q = testq, covariances = sdrobnorm(testy, lag = 4)^2 * testcorrelations), stepFit(testy, family = "mDependentPS", q = testq, filter = testfilter)) expect_identical(stepFit(testy, family = "mDependentPS", q = testq, covariances = 1.1^2 * testcorrelations), stepFit(testy, family = "mDependentPS", q = testq, filter = testfilter, sd = 1.1)) }) test_that("family 'jsmurf' works", { testn <- 70L testy <- rnorm(testn) testq <- rep(3, testn) testfilter <- lowpassFilter::lowpassFilter(type = "bessel", param = list(pole = 4L, cutoff = 0.1), sr = 1e4) expect_error(stepFit(family = "jsmurf", q = testq, filter = testfilter)) expect_identical(stepFit(testy, family = "jsmurf", q = testq, filter = testfilter), stepFit(testy, x = 1:testn, x0 = 0, family = "jsmurf", intervalSystem = "dyaLen", lengths = 2^(4:6), q = testq, confband = FALSE, jumpint = FALSE, filter = testfilter)) expect_error(stepFit(as.integer(testy), family = "jsmurf", q = testq, filter = testfilter)) expect_error(stepFit(c(testy, "s"), family = "jsmurf", q = testq, filter = testfilter)) expect_error(stepFit(c(rnorm(10), NA), family = "jsmurf", q = testq, filter = testfilter)) s <- stepFit(testy, q = testq, family = "jsmurf", confband = TRUE, filter = testfilter) bounds <- computeBounds(testy, q = testq, family = "jsmurf", filter = testfilter) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s <- stepFit(testy, q = testq, family = "jsmurfPS", confband = TRUE, filter = testfilter) bounds <- computeBounds(testy, q = testq, family = "jsmurfPS", filter = testfilter) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s <- stepFit(testy, q = testq, family = "jsmurfLR", confband = TRUE, filter = testfilter) bounds <- computeBounds(testy, q = testq, family = "jsmurfLR", filter = testfilter) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostGauss, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostGauss, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("family 'hjsmurf' works", { testn <- 70L testy <- rnorm(testn) testq <- rep(3, testn) testfilter <- lowpassFilter::lowpassFilter(type = "bessel", param = list(pole = 4L, cutoff = 0.1), sr = 1e4) expect_error(stepFit(family = "hjsmurf", q = testq, filter = testfilter)) expect_identical(stepFit(testy, family = "hjsmurf", q = testq, filter = testfilter), stepFit(testy, x = 1:testn, x0 = 0, family = "hjsmurf", intervalSystem = "dyaLen", lengths = 2^(4:6), q = testq, confband = FALSE, jumpint = FALSE, filter = testfilter)) expect_error(stepFit(as.integer(testy), family = "hjsmurf", q = testq, filter = testfilter)) expect_error(stepFit(c(testy, "s"), family = "hjsmurf", q = testq, filter = testfilter)) expect_error(stepFit(c(rnorm(10), NA), family = "hjsmurf", q = testq, filter = testfilter)) s <- stepFit(testy, q = testq, family = "hjsmurf", confband = TRUE, filter = testfilter) bounds <- computeBounds(testy, q = testq, family = "hjsmurf", filter = testfilter) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s <- stepFit(testy, q = testq, family = "hjsmurfSPS", confband = TRUE, filter = testfilter) bounds <- computeBounds(testy, q = testq, family = "hjsmurfSPS", filter = testfilter) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) s <- stepFit(testy, q = testq, family = "hjsmurfLR", confband = TRUE, filter = testfilter) bounds <- computeBounds(testy, q = testq, family = "hjsmurfLR", filter = testfilter) testx <- 1:testn testConfInt(s = s, b = bounds, x = testx, n = testn, tolerance = 1e-9) testFeasible(s = s, b = bounds, tolerance = 1e-9) testValues(s = s, b = bounds, localEst = localEstGauss, y = testy, tolerance = 1e-9) testCosts(s = s, y = testy, localCost = localCostHsmuce, tolerance = 1e-9) testOptimality(s = s, b = bounds, y = testy, localCost = localCostHsmuce, localEst = localEstGauss, tolerance = 1e-9) testBand(s = s, b = bounds, x = testx, tolerance = 1e-9) }) test_that("families LR and 2Param lead to errors", { testn <- 27L testy <- rnorm(testn) testfilter <- lowpassFilter::lowpassFilter(type = "bessel", param = list(pole = 4L, cutoff = 0.1), sr = 1e4) expect_error(stepFit(testy, family = "LR", q = rep(1, testn), filter = testfilter)) expect_error(stepFit(testy, family = "2Param", q = rep(1, testn), filter = testfilter)) })