tests/testthat/test-test_pse.R

In jrosen48/konfound: Quantify the Robustness of Causal Inferences

context("Checking test_pse")

test_that("test_pse handles non-positive standard error correctly", {
    # Expect an error to be thrown with the specific message when std_err is zero or negative
    expect_error(
        test_pse(est_eff = 0.5, std_err = 0, n_obs = 6174, n_covariates = 3, 
                 eff_thr = 0.1, sdx = 0.22, sdy = 1, R2 = 0.3, to_return = "full"),
        "Did not run! Standard error needs\n                              to be greater than zero."
    )
})



test_that("test_pse error handling for input validation", {
    # Testing that the function stops with an error message when standard deviation of x is zero or negative
    expect_error(
        test_pse(est_eff = 0.5, std_err = 0.056, n_obs = 6174, n_covariates = 3, 
                 eff_thr = 0.1, sdx = 0, sdy = 1, R2 = 0.3, to_return = "full"),
        "Did not run! Standard deviation of\n                          x needs to be greater than zero."
    )
})

test_that("test_pse error handling for input validation", {
    # Testing that the function stops with an error message when standard deviation of y is zero or negative
    expect_error(
        test_pse(est_eff = 0.5, std_err = 0.056, n_obs = 6174, n_covariates = 3, 
                 eff_thr = 0.1, sdx = 0.22, sdy = 0, R2 = 0.3, to_return = "full"),
        "Did not run! Standard deviation of\n                          y needs to be greater than zero."
    )
})

test_that("test_pse error handling for input validation", {
    # Testing that the function stops with an error message when there are too few observations
    expect_error(
        test_pse(est_eff = 0.5, std_err = 0.056, n_obs = 5, n_covariates = 3, 
                 eff_thr = 0.1, sdx = 0.22, sdy = 1, R2 = 0.3, to_return = "full"),
        "Did not run! There are too few observations relative\n           to the number of observations and covariates.\n           Please specify a less complex model to use KonFound-It."
    )
})

test_that("test_pse error handling for input validation", {
    # Testing that the function stops with an error message when R2 is not within the valid range (0,1)
    expect_error(
        test_pse(est_eff = 0.5, std_err = 0.056, n_obs = 6174, n_covariates = 3, 
                 eff_thr = 0.1, sdx = 0.22, sdy = 1, R2 = -0.1, to_return = "full"),
        "R2 needs to be greater than zero."
    )
})
test_that("test_pse error handling for input validation", {
    expect_error(
        test_pse(est_eff = 0.5, std_err = 0.056, n_obs = 6174, n_covariates = 3, 
                 eff_thr = 0.1, sdx = 0.22, sdy = 1, R2 = 1.5, to_return = "full"),
        "R2 needs to be less than one."
    )
})

test_that("test_pse error handling for input validation", {
    # Testing that the function stops with an error message when Rxz^2 calculation leads to a non-positive result
    expect_error(
        test_pse(est_eff = 0.5, std_err = 0.056, n_obs = 7, n_covariates = 3, 
                 eff_thr = 0.1, sdx = 0.22, sdy = 1, R2 = 0.999, to_return = "full"),
        "Did not run! Entered values produced Rxz^2 <=0,\n           consider adding more significant digits to your entered values.",
        fixed = TRUE 
    )
})

test_that("test_pse print output is correct", {
    # Setup for the test
    est_eff <- 0.5
    std_err <- 0.056
    n_obs <- 6174
    n_covariates <- 3
    eff_thr <- 0.1
    sdx <- 0.22
    sdy <- 1
    R2 <- 0.3
    to_return <- "print"
    
    # Expected string
    expected_output <- c(
        "This function calculates the correlations associated with the confound that ",
        "generate an estimated effect that is approximately equal to the threshold ",
        "while preserving the standard error.",
        "",
        "The correlation between X and CV is 0.214, and the correlation between",
        "Y and CV is 0.313.",
        "",
        "Conditional on the covariates, the correlation between X and CV is 0.248,",
        "and the correlation between Y and CV is 0.372.",
        "",
        "Including such a CV, the coefficient changes to 0.097, with standard error",
        "of 0.054.",
        "",
        "Use to_return = \"raw_output\" to see more specific results."
    )
    
    # Capture the output of the print statement
    output <- capture.output(
        test_pse(est_eff, std_err, n_obs, n_covariates, eff_thr, sdx, sdy, R2, to_return)
    )
    
    # Convert output array into a single string for easier grepl checking
    # output_string <- paste(output, collapse = "\n")
    
    # Check if the output is as expected
    expect_equal(output, expected_output)
})