Composite quantile periodogram for spectral analysis

Abstract

We propose a new type of periodogram for identifying hidden frequencies and providing a better understanding of the frequency behaviour. The quantile periodogram by Li () provides richer information on the frequency of signal than a single estimation of the mean frequency does. However, it is difficult to find a specific quantile that identifies hidden frequencies. In this study, we consider a weighted linear combination of quantile periodograms, termed 'composite quantile periodogram'. It is completely data adaptive and does not require prior knowledge of the signal. Simulation results and real-data example demonstrate significant improvement in the quality of the periodogram.