STARResult.likelihood_ratio_test

STARResult.likelihood_ratio_test(unrestricted)

Refuse the test: the threshold is unidentified under the null.

Parameters:

unrestricted (_ComparisonMixin) – Ignored.

Return type:

_LikelihoodRatioTest

Returns:

Never returns.

Raises:

SpecificationError – Always. Under linearity the threshold and the delay drop out of the likelihood entirely, so the likelihood ratio has no chi-squared limit and the nuisance parameters are present only under the alternative – the Davies problem. The correct procedure is a bootstrap or a supremum-type test (Hansen 1996), not a chi-squared tail. Returning a well-formed and wrong p-value here would be worse than declining.