Create and explore imaginative 3D worlds with our new product Dreamfield (opens in a new tab)

Freedman–Lane test validity under heteroscedasticity

Determine the level of the Freedman–Lane permutation test for interaction blocks in the componentwise distance-based regression model for triangular fuzzy data when the residuals are heteroscedastic.

Background

The paper proposes a Freedman–Lane permutation test for testing blocks of interaction regressors in the componentwise distance-based regression model for triangular fuzzy responses. The test relies on exchangeability of the residual vectors across units, an assumption that excludes heteroscedasticity depending on covariates.

A plasmode simulation assessed the test only for two interaction blocks under a correctly specified model with exchangeable residuals. The authors explicitly state that the test’s level under heteroscedasticity remains unstudied, leaving its inferential validity in that setting unresolved.

References

The Freedman–Lane test is approximate; in the plasmode experiment its size was between 0.025 and 0.055, but it was assessed only for the two blocks of that design, under a correctly specified model with exchangeable residuals, and its level under heteroscedasticity was not studied.