Nonnested hypothesis testing in the class of varying dispersion beta regressions

作者：Cribari Neto Francisco^*; Lucena Sadraque E F

来源：Journal of Applied Statistics, 2015, 42(5): 967-985.

DOI：10.1080/02664763.2014.993368

• bootstrap
• beta regression
• nonnested hypothesis testing
• nonnested models

Oftentimes practitioners have at their disposal two or more competing models with different parametric structures. Whenever each model cannot be obtained as a particular case of the remaining models through a set of parametric restrictions the models are said to be nonnested. Tests that can be used to select a model from a set of nonnested linear regression models are available in the literature. Particularly, useful tests are the J and MJ tests. In this paper, we extend these two tests to the class of beta regression models, which is useful for modeling responses that assume values in the standard unit interval,
[GRAPHICS]
. We report Monte Carlo evidence on the finite sample behavior of the tests. Bootstrap-based testing inference is also considered. Overall, the best performing test is the bootstrap MJ test. Two empirical applications are presented and discussed.

• 出版日期2015-5-4