ABSTRACT
In this chapter we survey some new methods of time-series analysis that are good at detecting low-dimensional chaos as well as other nonlinear-ities that are useful for short-term prediction but might be missed by many more conventional statistical methods. The methods are extended to handle the problem of detecting nonlinear correlation between two different series above and beyond linear relationships. After development of these methods in Section Two, they are applied in Section Three. Section Three also discusses statistical issues raised by evaluating superior predictive performance of stock market technicians. Section Four discusses very speculative pathways by which fluctuations in economic activity arise.
