ABSTRACT
Methodologies for analyzing multiple time series data are discussed. These have applications from pairs trading to portfolio optimization to trading in multiple markets. A key feature of this chapter is the discussion of multivariate dimension-reduction techniques that play an important role in large-scale data modeling. Because the data is chronological, the commonly used techniques such as principal component analysis may not be valid; we discuss the concept of reduced-rank regression that has become widely used in practice. Modeling of multiple non-stationary series such as stock prices is discussed. The concept of co-integration, commonality, co-movement are covered in detail with illustrative examples. Recent research on Multivariate GARCH models is also discussed with applications in foreign exchange market.
