No abstract

On a high-frequency scale, financial time series are not homogeneous, therefore standard correlation measures can not be directly applied to the raw data. To deal with this problem the time series have to be either homogenized through interpolation or methods that can handle raw non-synchronous time series need to be employed. This paper compares two traditional methods that use interpolation with an alternative method applied directly to the actual time series. The three methods are tested on simulated data and actual trades time series. The temporal evolution of the correlation matrix is revealed through the analysis of the full correlation matrix and of the Minimum Spanning Tree representation. To perform the analysis we implement several measures from the theory of random weighted networks.

On a high-frequency scale the time series are not homogeneous, therefore standard correlation measures can not be directly applied to the raw data. There are two ways to deal with this problem. The time series can be homogenised through an interpolation method [1] (linear or previous tick) and then the Pearson correlation statistic computed. Recently, methods that can handle raw non-synchronous time series have been developed [2,4]. This paper compares two traditional methods that use interpolation with an alternative method applied directly to the actual time series.

In this paper we implement a Fourier method to estimate high frequency correlation matrices from small data sets. The Fourier estimates are shown to be considerably less noisy than the standard Pearson correlation measure and thus capable of detecting subtle changes in correlation matrices with just a month of data. The evolution of correlation at different time scales is analysed from the full correlation matrix and its Minimum Spanning Tree representation. The analysis is performed by implementing measures from the theory of random weighted networks.