|
PureBytes Links
Trading Reference Links
|
Paul
Your criticism of Mark's analysis seems to be partially based on the fact
that there are "only" 13 annual samples. Would the statistical analysis be
more complete if you broke his study down into 6-month or 3-month segments
to provide more samples? Just from eyeballing the equity curve, it is
obvious there are some significant drawdowns. Perhaps the drawdowns of the
system should be more the focus of your stance than the StdDev of annual
returns? Also, if I remember correctly, the purpose of Mark's study was to
examine the 2 different betsizing methods and the system chosen was
admittedly arbitrary.
Kent
-----Original Message-----
From: Paul H. Lasky <phlasky@xxxxxxxxxxxxx>
To: Omega-List@xxxxxxxxxx <Omega-List@xxxxxxxxxx>
Date: Sunday, September 12, 1999 7:01 PM
Subject: Fixed Ratio and other Pyramiding Schemes.
standard error of the average yearly return with S.E. = Std.Dev/Sqr(13). The
Std. Deviation
is calculated from each of the 13 annual sample returns. The results are:
A note about the dependent t test. Because Mark computed equity curves
for each scheme on the SAME commodity data, they can be compared using a
more sensitive statistical test called the dependent t test. Subtract the
yearly returns of each pair and compute the standard deviation of the
difference over the 13 differences representated in Mark's data. This test
4) Mark presented us with 13 independent statistical samples for the
13 years of annual results for the pyramid schemes. Valid statisical
analysis can be performed with 13 samples. But Mark's contrived example
containes only a single sample.You can't compute standard deviations with a
single sample. Valid statistical analysis cannot be performed on one sample.
( See any first year course in statistics.)zx
|