An Approach to Modeling on Financial Time Series Data with Regime Shifts

Abstract

This paper proposes a new method to analyze time series data with regime shifts and makes the following three contributions: (1) it suggests an exponential weighted estimation algorithm for autoregressive model with time varying coefficients, (2) it gives a visualization technique of structural change points and an outlier measure based on the Mahalanobis distance and (3) it illustrates that our method works for hedge fund return data and high frequency FX data.