To get a statistically correct result, 10x is the minimum set that
makes any sense otherwise I have to go to t-values and the error of
the estimate goes out of sight.
I like to have some precision in my testing... ;-)
Godfrey
degree in mathematics, specialization in statistics
On Sep 8, 2005, at 5:02 AM, Shel Belinkoff wrote:
For some you may be right. For others, something a bit more
precise is
desirable. Speaking only for myself, something a bit more precise
than 10
mb/sec v 20mb/sec is desirable. More precision, numbers closer to
actual,
are more meaningful. However, I'm not sure - in fact I know - that
I'd not
run the test 10X per card that I'm testing. Eight sounds about
right ;-))
Shel
[Original Message]
From: keith_w
Subject: Re: SD Ultra II - testing ...
Ten times? That seems like serious overkill to me.
The effort expended is MORE than needed. This is not the national
bureau
of standards level of testing being required. If you're off by 2
seconds, or 3 or 5, do you beiieve it will really matter to anyone?
When someone asks the data transfer rate for a given card, they
mean is
it 10 mb/sec or 20 or 30. Not 11 vs. 10.
[NB: The above sentence is not meant to reflect actual numbers,
but is
merely illustrative, to get the point across...] <g>
keith
