Title: Departures from symmetry and testing for heteroskedasticity in nonparametric regression
Authors: Daniel Henderson - University of Alabama (United States) [presenting]
Alice Sheehan - University of Alabama (United States)
Abstract: A new conditional moment test for heteroskedasticity in nonparametric regression models is proposed. The test statistic uses kernel based estimation. Our test builds on previous work, but relaxes the assumption that the density function have compact support. This eliminates the need to trim small values of the data to ensure the test statistic is well behaved. We then introduce a naive bootstrap method as an alternative to the wild bootstrap method previously employed. We show asymptotic equivalence of our test statistic and bootstrap method. A Monte Carlo simulation is provided, with critical values obtained from both the wild bootstrap and a naive bootstrap, to assess the finite sample performance of the proposed test as well as the bootstrap method.