Bruce Weaver <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>...
> Perhaps because the way the problem was described, persons A 
> and B are flipping the *same* coin, not two different coins?

I understood the circumstance. That might affect the probability
of success, but I don't see how to introduce much dependence.

Could you describe a way in which you believe that B's 3rd flip
could be dependent on A's 3rd flip? A's 5 flip? B's first flip?

I can see only two kinds of dependence. One would be extremely
small (the coin wears as you use it, so p may slowly change over
time, which introduces a dependence of a kind). 

The other could be less small, but could be dramatically reduced by
appropriate throwing protocols (as Persi Diaconis and others have
shown,
each coin flip isn't absolutely independent of the face that is
uppermost
when you flip it, so if you always pick up the coin and place it for
the
next flip in the same way, so the face that's uppermost when you flip
it
next time is perfectly dependent on the face that was uppermost when
it
landed last time, there is a potential for a bit of serial
dependence).

Both these effects will be small relative to the noise in your
estimates
unless the number of tosses is very, very large indeed.

Neither of these effects really come into what most people mean when
they talk of "dependent samples", where they'd posit some kind of
relationship between A's i-th toss and B's i-th toss.

Glen
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to