A model sufficiency test using permutation entropy

Using the ordinal-pattern concept in permutation entropy, we propose a model sufficiency test to study given model's point prediction accuracy. Compared with some classical tests, such as Broock et al.'s (1996) test, our proposal does not require sufficient eliminate all structures exhibited estimated residuals. When innovations investigated data's underlying dynamics show certain structure, higher moment serial dependence, can lead erroneous conclusions about of predictors. Due structured innovations, inconsistency between tests and accuracy criteria occur. Our fills this incoherence evaluation approaches remains valid when process has nonwhite additive innovation.