A0391
Title: Kernel mode-based varying coefficient models with nonstationary regressors
Authors: Tao Wang - University of Victoria (Canada) [presenting]
Abstract: Varying coefficient models are estimated on the basis of mode value using a kernel objective function, where the regressors are generated by multivariate unit root processes but can also be stationary. Such a kernel model-based estimation is demonstrated to be more robust and efficient than least squares estimation for data with outliers or heavy-tailed distributions, and it does not lose any efficiency when the data follow a normal distribution. A local linear approximation scheme is developed to estimate the varying coefficient function. Under mild regularity conditions, the asymptotic normality of the resultant estimators for both the unknown varying coefficient function and its derivative function is established. It is shown that the nonparametric estimator of the varying coefficient function with nonstationary regressors converges faster than the estimator with stationary regressors. In order to achieve estimation optimality in the sense of minimizing the asymptotic mean squared error, a kernel mode-based two-step estimation procedure is then suggested. The finite sample performance of the developed estimator is illustrated through three Monte Carlo simulations as well as a real data application on evaluating credit rationing in the United States credit market.
