Wavelets for High Frequency Financial Time Series

Keywords: Wavelets Transforms; High Frequency Stock Returns Series; Nonparametric Statistics Inference

Abstract: We analyze a Nikkei stock index series of minute-by-minute observed returns and study what wavelets transforms suggest in terms of volatility features underlying the sampled data and other derived series. One of the goals is to use wavelets as a pre-processing de-noising tool, so to help in identifying, estimating and predicting the volatility function. We also aim to investigate some temporal aggregation results obtained for financial time series. We thus conduct our experiments with an interest toward both a statistical inference perspective and an empirical finance viewpoint. We not only put the emphasis on the high frequency data sets but also report evidence about how new nonparametric statistics procedures such as wavelets help in uncovering features in the data generating process which were not found from the observed noisy returns series.