Letters to Technical Analysis
Dear Dr. Warren,
The purpose of this letter is to request your help, as the author, to obtain a working version of the
program for Maximum Entropy Forecasting which appeared in the December, 1984 issue of Technical
Analysis of Stocks & Commodities.
I have become interested in using a technique such as maximum entropy to evaluate cross-over points to
try to eliminate whip-sawing. The initial analysis section published in the July/ August issue seems to be
functioning with the corrections. However, the later section for forecasting from the December issue
appears to have a glitch. Terms L%, NU, and DL are used in line 1160 without having been defined and
the net result is a division by zero error.
Edward S. Yorkison
Thank you for your recent letter pointing out a problem with the Maximum Entropy Forecasting routine
published in the December 1984 issue of Technical Analysis of Stocks & Commodities magazine. I have
not been able to respond earlier due to vacation schedules and pressing business.
Line 1160 of the Forecasting code is a special code adaptation for use within the CompuTrac system and
is indeed not generic code. We have since determined that most CompuTrac users would prefer using our
diskette code versions, and will in the future publish more generic Basic code (the code for the trend
following method in the June 1985 issue of Technical Analysis of Stocks & Commodities will be
published in generic form for example).
The problem in line 1160 can be solved by replacing the first part of line 1160 which currently appears