You should talk to your friendly neighborhood statistician, but here
are my $.02:
By "reproduce within a tenth" let's say you mean "the standard
deviation from the average is +/- 0.1 sec". That would put ~ 68% of
runs in this window. Assuming only that all runs are normally
distributed about the mean, it is a safe bet that 1 in 25 would be 3 SD
(0.3 sec) different.
If by "reproduce within a tenth" you mean _every_ run except the
oddball is +/- 0.1 sec ... this would require (arguably) 68% (or 90%)
of runs to be +/- 0.05 (0.03) sec of the mean, again assuming only
normal distribution of error. That is pretty hard to swallow, hence my
first assumption (which only needs to be chewed very well).
This argument is independent of the source of the variation (lights,
clock, lead foot, warm tires, butterflies in South America).
sb
On Jun 1, 2004, at 10:55 PM, Richard Urschel wrote:
> On Sat, 29 May 2004 18:09:22 -0700 (PDT) Ms Katie Kelly
> <aceontour@yahoo.com> wrote:
>> If the timers are not accurate, then how do you know
>> that you're within .1 in 25 or any amount of runs?
>> They could be wrong on every run.
>
> First, let me point out a couple of things. I did not claim nor agrue
> that the timers are, in fact, inaccurate.
> I asked how accurate they are, and provided a basis for suspecting
> they might have an occasional problem with consistency in tripping. My
> basis was neither compelling nor conclusive, nor did I expect any one
> to perceive it as an attempt at such. It was an academic question.
> Sometimes a cigar is just a cigar.
|