Ack, wrote this right before John sent his email.....

Sarbajit, et al,
Technically it would be a binomial distribution of
some ugly sort. There would be some skew, or at least a tail cut off in one
direction earlier than the other, but for many purposes a normal distribution
approximates the binomial distribution. 

The probability of getting a
glass with 0 molecules in it, if there was random distribution, would be quite
small. Based on the numbers given, the change of drawing any given molecule, if
drawing one at a time, would be 1/8x10^21. However, you are drawing 8x10^24
molecules at a time. So, on average you should have 1000 molecules per glass.
However, the probability of getting any specific number of molecules per glass
is quite small. If we use the approximation of the normal curve, and this is
just the type of situation in which that is ungodly helpful, the standard
deviation of the number of molecules in any glass is (I hope) 31.6 molecules.
So 99% of the glasses you draw will have between 918 and 1082 molecules from
the original glass. 99.99% of the sample glasses will have between 877 and 1123
molecules from the original glass. 

I really hope that is right, I think
it is,

Eric

P.S. The standard deviation of the binomial
distribution = the square root of (the number of trials x the probability of a
success x the probability of failure). In this case we have (8*10^24) trials,
with a probability of success = (1/8x10^21),  and a probability of failure
= (1-the prob of success). The z-scores marking off 99% of the distribution are
plus and minus 2.58, for 99.99%, z=3.90.

P.P.S. I might be horribly off
because of how low the probability of a success is. I think the ridiculously
high number of trials mitigates that, but I'm not sure how much. 


On
Tue, Apr 27, 2010 02:22 PM, sarbajit roy <[email protected]>
wrote:
>
>
>
>Hi Steve
>
>" The chances of
>drawing a glass without any marked molecules is 1/1000, supporting ES's
>claim."
>
>I don't think the maths works quite that way. Some glasses would
>have exactly 1000 molecules, some would have 1000 -/+ 1, or 2 .. 
>-/+999.  Presuming that the distribution is a "normal" distribution, there
>would be an exceedingly small probability of getting a glass with zero marked
>molecules.
>
>Furthermore since there is the equally remote probability
>that a single glass would contain all the marked molecules (just like we
>started out with), the distribution would be skewed away from a normal
>one..
>
>This is just an off the cuff observation. I could brush up my
>prob-stats if reqd (and eat humble pie if wrong).
>
>


>>On Tue, Apr 27, 2010 at 9:35 AM, Steve Smith <<#>> wrote:
>
>
>
>
>  
>
>>
>Nick -
>
>
>I read it through before seeing your retraction.  As you may recognize
>by now, your fallacy is probably not a consequence of your being an
>English (Psychology?) Major but actually just not reading the statement
>of the problem carefully enough.   The 10^24 (molecules) vs the 10^21
>glasses (cups?)  might be about right and your math is good (1000
>molecules per glass on average)... but the conclusion (1/1000 chance of
>drawing a glass with a marked molecule) is reversed.   The chances of
>drawing a glass without any marked molecules is 1/1000, supporting ES's
>claim.   
>
>
>I'd say you did good (right up to that premature send thingy) for an
>English Major.
>
>
>I read ES's "What is Life" years ago and was deeply inspired by it's
>directness and simplicity (and lack of jargon) and timeliness (1949?)
>well before much was done to tie life to information theory.   I look
>forward to your continued "book reports".
>
>
>- Steve
>
>
>
>
>
>
>
============================================================
>FRIAM Applied
>Complexity Group listserv
>Meets Fridays 9a-11:30 at cafe at St. John's
>College
>lectures, archives, unsubscribe, maps at
>http://www.friam.org
>
>

Eric Charles

Professional
Student and
Assistant Professor of Psychology
Penn State
University
Altoona, PA 16601



Eric Charles

Professional Student and
Assistant Professor of Psychology
Penn State University
Altoona, PA 16601


============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

Reply via email to