Statistics Seminar（2014-02）

Topic:Significance Testing in Nonparametric Autoregression

Speaker:Xiaojun Song, Universidad Carlos III de Madrid

Time:Thursday, 13 February, 10:30-11:30

Location:Room 217, Guanghua Building 2

Abstract：We propose significance tests in nonparametric autoregression. Under the null, forecast of any nonlinear autoregression of order p is unaffected by considering any extra lagged value. A necessary and sufficient condition, which forms a basis for the tests, is that the residuals of the p-th order nonparametric autoregression are uncorrelated with any measurable function of the lagged variables. The test statistic is based on Fourier transform of the autocorrelation function of the nonparametric residuals and functions of the lagged values. The tests are implemented with the assistance of a bootstrap technique. We illustrate the practical performance of the test by means of simulations and an empirical application of stock returns.