Title: An asymptotically efficient test for functional coefficient models
Authors: Xingtong Zhang - Cornell University (United States) [presenting]
Abstract: Functional coefficient models have abilities to capture non-linearity and heteroscedasticity by allowing coefficients to be governed by some variables. To test the model specifications, there are two general approaches: the generalized likelihood ratio (GLR) test proposed and loss function approach. Despite enjoying appealing features such as Wilks phenomena, they both rely on nonparametric convergence rate and thus suffer from curse of dimensionality. We propose a root-T consistent test using Fourier transforms. The new test is asymptotically more efficient than both GLR test and loss function approach. Because of its parametric convergence rate, our test is free from dimension of nonparametric smoothing.