Inference of weighted \(V\)-statistics for nonstationary time series and its applications (Q2448724)

The article deals with statistical inference in piece-wise locally stationary (PLS), potentially nonlinear, time series models. Central and non-central limit theorems are proved for weighted \(V\)-statistics of PLS time series data, both in the nondegenerate and in the degenerate case. To this end, under regularity assumptions, a Fourier integral representation of the \(V\)-statistic kernel is employed. This leads to a mathematically tractable structure of the degenerate and the nondegenerate part appearing in the Hoeffding-based decomposition of the \(V\)-statistic. Applications of the main results comprise asymptotic distributional theory for quadratic forms of PLS processes, point-wise central limit theorems for certain nonparametric estimators of time series parameter functions, and asymptotic distributional results in the context of a spectral analysis based on the (empirical) periodogram of the PLS time series.