Struct SpecificationTests 

pub struct SpecificationTests;

Specification tests for regression models

Implementations

impl SpecificationTests

pub fn white_test( residuals: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>, x: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 2]>>, ) -> Result<(f64, f64, usize), String>

White’s Test for Heteroskedasticity

Tests H₀: Homoskedasticity (constant variance) vs H₁: Heteroskedasticity

Arguments

• residuals - Residuals from OLS regression
• x - Design matrix (n × k)

Returns

Tuple of (LM_statistic, p_value, degrees_of_freedom)

Interpretation

• If p < 0.05: Reject H₀, heteroskedasticity is present (use robust SE)
• If p > 0.05: Fail to reject H₀, homoskedasticity is plausible

use greeners_diagnostics::specification_tests::SpecificationTests;
use ndarray::{Array1, Array2};

// ... the rest of your original example here ...
let (lm_stat, p_value, df) = SpecificationTests::white_test(&residuals, &x)?;

pub fn reset_test( y: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>, x: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 2]>>, fitted_values: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>, power: usize, ) -> Result<(f64, f64, usize, usize), String>

Ramsey RESET Test for Functional Form Misspecification

Tests H₀: Model is correctly specified vs H₁: Functional form misspecification

Arguments

• y - Dependent variable
• x - Design matrix (n × k)
• fitted_values - Fitted values from original regression
• power - Maximum power of fitted values to include (typically 2, 3, or 4)

Returns

Tuple of (F_statistic, p_value, df_num, df_denom)

Interpretation

• If p < 0.05: Reject H₀, functional form misspecification detected
• If p > 0.05: Fail to reject H₀, functional form appears adequate

use greeners_diagnostics::specification_tests::SpecificationTests;
use ndarray::{Array1, Array2};

let (f_stat, p_value, _, _) = SpecificationTests::reset_test(&y, &x, &fitted, 2)?;

pub fn breusch_godfrey_test( residuals: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>, x: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 2]>>, lags: usize, ) -> Result<(f64, f64, usize), String>

Breusch-Godfrey Test for Autocorrelation

Tests H₀: No autocorrelation up to lag p vs H₁: Autocorrelation present

Arguments

• residuals - Residuals from OLS regression
• x - Design matrix (n × k)
• lags - Number of lags to test (typically 1 for AR(1))

Returns

Tuple of (LM_statistic, p_value, degrees_of_freedom)

Interpretation

• If p < 0.05: Reject H₀, autocorrelation detected
• If p > 0.05: Fail to reject H₀, no evidence of autocorrelation

Examples

use greeners_diagnostics::specification_tests::SpecificationTests;
use ndarray::{Array1, Array2};

// The visible example starts here:
let (lm_stat, p_value, df) = SpecificationTests::breusch_godfrey_test(&residuals, &x, 1)?;

pub fn goldfeld_quandt_test( residuals: &ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>, split_fraction: f64, ) -> Result<(f64, f64, usize, usize), String>

Goldfeld-Quandt Test for Heteroskedasticity

Tests H₀: Homoskedasticity vs H₁: Variance increases with ordering variable

Arguments

• residuals - Residuals from OLS regression (should be ordered by suspected variable)
• split_fraction - Fraction of middle observations to drop (typically 0.2 to 0.33)

Returns

Tuple of (F_statistic, p_value, df1, df2)

Interpretation

• If p < 0.05: Reject H₀, heteroskedasticity detected
• If p > 0.05: Fail to reject H₀, homoskedasticity plausible

pub fn print_test_result( test_name: &str, statistic: f64, p_value: f64, null_hypothesis: &str, alternative_hypothesis: &str, )

Pretty print specification test results