[OPE-L:5528] [PAUL C] Re: Statistical conservation of value

Gerald Levy (glevy@pratt.edu)
Thu, 25 Sep 1997 06:16:48 -0700 (PDT)

[ show plain text ]

Bohr was right, mass/energy conservation only holds on
large enough scales. We have made the same analogy
with nonconservation of energy in 'Value's law values Metric' webpublished
as
http://www.cs.strath.ac.uk/Contrib/wpc/reports/metric/metric.html

Whilst I consider Ochoa and Petrovic to have been mistaken in
saying that one can get spurious correlations between values and prices,
I think there is some merit in the argument that the appropriate
metric to minimise is not the squared error - which is what standard
regression and correlation techniques attempt to do.

Various possible metrics could be used:

Euclidean distance
Angle between vectors
Inner product of the vectors
Correlation
Mean absolute error

In the paper cited we argue that commodity space is non-euclidean
and that a modified manhattan metric is the appropriate one to use.
This would imply that statistical estimation techniques applied
to commodity prices should be designed to minimise the mean absolute
error.