tests/testthat/test-eive-formula-otherx.R

In eive: An Algorithm for Reducing Errors-in-Variable Bias in Simple and Multiple Linear Regressions

library(testthat)
library(eive)

test_that("eive.cga classical example using formula with other X1 and X2", {
    tol <- 0.01

    euclidean <- function(u, v) {
        return(sqrt(sum((u - v)^2)))
    }

    require("eive")

    set.seed(12345)

    n <- 30

    clean_x <- rnorm(n, mean = 10, sd = sqrt(7))

    delta_x <- rnorm(n, mean = 0, sd = sqrt(3))
    other_x1 <- rnorm(n, mean = 0, sd = sqrt(3))
    other_x2 <- rnorm(n, mean = 0, sd = sqrt(3))

    e <- rnorm(n, mean = 0, sd = sqrt(5))

    y <- 20 + 10 * clean_x + 10 * other_x1 + 10 * other_x2 + e

    dirty_x <- clean_x + delta_x

    mydata <- data.frame(
        y = y,
        dirtyx = dirty_x,
        otherx1 = other_x1,
        otherx2 = other_x2
    )

    result <- eive.cga.formula(
        formula = y ~ dirtyx + otherx1 + otherx2,
        dirtyx.varname = "dirtyx",
        data = mydata,
        numdummies = 10
    )

    par_real <- c(20, 10, 10, 10)
    par_ols <- result$ols$coefficients
    par_eive <- result$eive$coefficients

    dist1 <- euclidean(par_real, par_ols)
    dist2 <- euclidean(par_real, par_eive)

    expect_true(dist2 < dist1)

    expect_equal(
        as.vector(par_eive),
        c(27.875, 8.876, 8.280, 9.624),
        tolerance = tol
    )

    expect_equal(
        as.vector(par_ols),
        c(65.868, 5.232, 9.960, 7.396),
        tolerance = tol
    )

    L <- length(result$measurementerror)
    for (i in 1:L) {
        expect_equal(result$proxy$residuals[i], result$measurementerror[i])
    }

    result_names <- names(result)
    expect_true("ols" %in% result_names)
    expect_true("eive" %in% result_names)
    expect_true("proxy" %in% result_names)
    expect_true("cleanedx" %in% result_names)
    expect_true("measurementerror" %in% result_names)
})





test_that("eive.cga classical example using formula with other X1 (single other)", {
    tol <- 0.01

    euclidean <- function(u, v) {
        return(sqrt(sum((u - v)^2)))
    }

    require("eive")

    set.seed(12345)

    n <- 30

    clean_x <- rnorm(n, mean = 10, sd = sqrt(7))

    delta_x <- rnorm(n, mean = 0, sd = sqrt(3))
    other_x1 <- rnorm(n, mean = 0, sd = sqrt(3))

    e <- rnorm(n, mean = 0, sd = sqrt(5))

    y <- 20 + 10 * clean_x + 10 * other_x1 + e

    dirty_x <- clean_x + delta_x

    mydata <- data.frame(
        y = y,
        dirtyx = dirty_x,
        otherx1 = other_x1
    )

    result <- eive.cga.formula(
        formula = y ~ dirtyx + otherx1,
        dirtyx.varname = "dirtyx",
        data = mydata,
        numdummies = 10
    )

    par_real <- c(20, 10, 10)
    par_ols <- result$ols$coefficients
    par_eive <- result$eive$coefficients
    dist1 <- euclidean(par_real, par_ols)
    dist2 <- euclidean(par_real, par_eive)

    expect_true(dist2 < dist1)

    expect_equal(
        as.vector(par_eive),
        c(23.453141, 9.070309, 11.589761),
        tolerance = tol
    )

    expect_equal(
        as.vector(par_ols),
        c(64.933586, 5.317319, 10.075051),
        tolerance = tol
    )

    L <- length(result$measurementerror)
    for (i in 1:L) {
        expect_equal(result$proxy$residuals[i], result$measurementerror[i])
    }

    result_names <- names(result)
    expect_true("ols" %in% result_names)
    expect_true("eive" %in% result_names)
    expect_true("proxy" %in% result_names)
    expect_true("cleanedx" %in% result_names)
    expect_true("measurementerror" %in% result_names)
})