John's probably got a better handle on it than I do. I generally go
overboard and go a magnitude up on the entire range rather than the real
range. Means when I do a test here at work I'd go up to 1000 occurrences
minimum on a % situation.
Kit
-----Original Message-----
From: Ryan Honeyman [mailto:
honeyman@...]
Sent: Tuesday, January 25, 2000 2:46 PM
To: '
eqbards@onelist.com'
Subject: RE: [eqbards] Deepwater Gobs
From: Ryan Honeyman <
honeyman@...>
> Just a brief word on statistics. If you want to get a reasonably good
> approximation of what % something drops, you need a number of tests one
> order of magnitude higher than the range. So if its 5% you need at least
> 200 tests for statistical accuracy. This isn't necessarily true for all
> randoms, but for psuedo-random numbers such as computers produce, it is
> generally the applicable rule.
I'll gladly defer this to you and John(tm). I'm weak at
statistics ;-) All I know is I've been really damn lucky
and really damn unlucky hehe.
Harm.
--------------------------- ONElist Sponsor ----------------------------
Accurate impartial advice on everything from laptops to table saws.
<a href="
http://clickme.onelist.com/ad/Productopia ">Click Here</a>
------------------------------------------------------------------------
Please send submissions for the eqbards newsletter to
lol@...
with the subject submissions.