|
Abstract:
|
We investigate tests for periodicity based on a spectral analysis of a time series. We contrast between global and local spectral-based tests. A global test uses the spectral estimates at all sample points, whereas a local test uses a subset of the sample points available. Standard global tests for periodicity are often based on the assumption of a Gaussian IID error process. Using a smoothing spline approach, we extend the global test to the non-IID case. We compare this approach to a number of local tests for periodicity such as the local F test, a test commonly used in hearing sciences, and Thomson's multitaper F test. Using regression-based F tests, we demonstrate that asymptotic size and power calculations can be made for some of these tests. For all the tests, we also compare the size and power at finite sample sizes, under a number of different experimental conditions.
|
