Interpreting measured serial correlation in univariate time series analysis, with an example from the New York Stock Exchange

This paper draws attention to the divergence of the observed serial correlations, for a typical given time series, from the corresponding theoretical autocorrelation pattern of the underlying generating process, which is not just a matter of sampling variation, and which seems to be commonly overlooked when analysing data in finance. The present thesis is illustrated with examples from a case-study in the recent literature, where stock market transactions data for two distinct periods were analysed, and the behaviour of returns and various trading characteristics examined.