Testing for nonlinearity in time series: the method of surrogate data
Conference proceedings on Interpretation of time series from nonlinear mechanical systems
Genetic Programming Prediction of Stock Prices
Computational Economics
| Hi-index | 0.00 |
We review a complexity measure (&thgr;)and its statistical properties, then apply it to fourstock returns. Theta is a ratio of two correlationintegral estimates, one taken before and one aftershuffling the series to investigate. For randomprocesses &thgr; ≈ 1 while &thgr; approaches zerofor data with low complexity. Sixteen artificiallygenerated series with different dynamicalcharacteristics – each represented by three samplesizes – were employed to investigate &thgr;‘sstatistical properties. Its distribution approachesnormality as the sample size is increased. Whenapplied to stock returns, those computed at everyprice change proved less complex than lower frequencyone- and five-minute returns, implying thatinformation is being lost by increasingly lessfrequent sampling.