|
PureBytes Links
Trading Reference Links
|
Although this may seem a bit off topic, it may be very important to
those who use Excel (and some other programs) to statistically
analyze results from AB.
The following study reports that for certain data sets, Excel's
results can be off by as much as 8 orders of magnitude (ie, 1
million times too low or too high). Microsoft cynics may not be
surprised by this, but I find it shocking that this "flaw" is in
Excel "by design" according to authors of the study. Apparently
Excel uses a mathematical short cut for some statistical sums that
works in many cases but fails in other.
http://web.polmeth.ufl.edu/papers/99/altmcd99.pdf
b
|