Martingale-difference-divergence-based tests for goodness-of-fit in quantile models

Kai Xu,Fangxue Chen
DOI: https://doi.org/10.1016/j.jspi.2019.10.007
IF: 1.095
2020-01-01
Journal of Statistical Planning and Inference
Abstract:In this article, we use the martingale difference divergence (Shao and Zhang, 2014) to propose novel tests for the goodness-of-fit of linear quantile models. In the aspect of methodology, our proposals are applicable under heteroscedastic models, easy to implement, and free of user-defined regularization. In the aspect of theory, our analyses are of own interest as we need to deal with the residuals that are not differentiable with respect to the regression parameter. We investigate distribution theory for the test statistics under both the null and the local and fixed alternative sequences. A bootstrap method is proposed to approximate the critical values of the tests and its consistency is also justified. Numerical studies are used to illustrate the empirical merit of our proposed tests. (C) 2019 Elsevier B.V. An rights reserved.
What problem does this paper attempt to address?