|
PureBytes Links
Trading Reference Links
|
At 9:40 PM -0700 4/13/01, BobR wrote:
>Is there any easylanguage code that will determine how well a moving
>average represents price? Suppose for example you want to compare
>different types of averages and different lengths of averages and
>desire a numerical output for fitness of each average. How would you
>write a statistical indicator to compare the close or the (H+L)/2 to
>the moving average over a period of N bars or maxbars back?
You can start by knowing that the average will lag the price. That
will be most of the error. I would assume you want to remove that
source of error first and see what is left.
The lag of many common averages (Average, XAverage, etc.) is given by:
Lag = (Length - 1) / 2
More complex averages such as the T3Average and the JurikMA have more
complicated lag functions.
So next you might want to calculate the root-mean-square error -
something like this:
Avg = Average(Price, Length);
Lag = (Length - 1) / 2;
Error = Price[Lag] - Avg;
SqrErr = Error * Error;
MSqrErr = Average(SqrErr, Length2);
if MSqrErr > 0 then RMSE = SquareRoot(MSqrErr);
or the average absolute error - something like this:
Avg = Average(Price, Length);
Lag = (Length - 1) / 2;
Error = Price[Lag] - Avg;
AbsErr = AbsValue(Error);
MAbsErr = Average(AbsErr, Length2);
The choice will depend upon what you are trying to determine.
Bob Fulks
To unsubscribe from this group, send an email to:
realtraders-unsubscribe@xxxxxxxxxxxxxxx
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
|