A simple analogue of auto- and cross-correlation
by Clifford J. Sherry
If you trade commodities or stocks and if you expect to make a profit, you are making the tacit
assumption that you can predict the future price of your commodity or stock.
Traders often need to know if time series of commodity or stock prices are cyclic and, if they are, the
extent of the cycle. It is also important to know if two time series are interdependent. Interdependency
can be between two commodity prices, if you want to play the spreads, or a series of commodity prices
and some other time series, such as the Composite Index of Leading Economic Indicators. These
properties can be detected by using auto- and cross-correlation techniques, respectively. Unfortunately,
these techniques are often difficult to implement and may require a large, general-purpose computer to do
During the last 10-12 years, with a number of different colleagues, I have developed a number of
statistical techniques to deal with these and related questions. Since auto- and cross-correlation are
relatively complex and difficult to implement, I developed an analogous technique that can be done with
paper and pencil or a small computer.
The first step is to obtain a listing of the prices of the commodity or stock. If you are using prices, they
should be as recent as possible and should represent several hundred equally spaced data points (prices).
They should cover the same calendar period.