A regression line is one projected through the center of a series of data points. The gold volume in millions of ounces traded per day on the London Bullion Market (Above) has such a regression line shown in black.

The R^2 value of the regression line (.784) is a mathematical measure of the nearness of the points to the regression line with 1 being perfect (All points fall exactly on the line). The closer to 1 the R^2 value falls, the greater the predictive power of a regression line's extension into the future (Provided the dominant forces have not changed).

Clearly, the earlier data points are farther from the line than current points in the example above so we can qualitatively say that the current R^2 value is higher than the past data points. The writer has developed methods to examine quantitatively this kind of difference for better predictive capacity.

The above descending linear regression pattern strongly suggests that the LBMA daily gold volume has been falling and will continue to do so until a reaction event involving the price of gold occurs.

Testing for intervention

The R^2 value can be a useful test for intervention as the force application is almost always done via computer directed actions. An example of this is any long period where the trading in say, the DIVG (a derived metric) remains close to the regression line to a high R^2 value. Regression lines can also be curved. If the R^2 values are very high and the curve is a simple polynomial, intervention can't be ruled out.

Since intervention has been shown to be intermittent, there is a need for multi-phasic regression techniques to examine only those data intervals suspected of abnormal activity in order to validate its existence.

The preferred mechanism for intervention is through the output of an econo-physical computer algorithm and this out put tends to leave tracks. It is these computer tracks that I seek.