Re: Howard Hershey's Challenge of Sean Pitman's Assumptions

On Dec 17, 8:34 am, "R. Baldwin" <res0k...@xxxxxxxxxxxxxxxxxxxx>
wrote:
"Seanpit" <seanpitnos...@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message

news:e3b58c91-802c-4bb7-a552-6cb31660150b@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx



On Dec 16, 10:02 am, "R. Baldwin" <res0k...@xxxxxxxxxxxxxxxxxxxx>
wrote:
"Seanpit" <seanpitnos...@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message

news:1a8d31d0-a7da-46cc-8f0c-68e2433a6efd@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

On Dec 15, 11:08 am, "R. Baldwin" <res0k...@xxxxxxxxxxxxxxxxxxxx>
wrote:

Neither is "minimum actual gap size".

That's statistically mistaken. Minimum likely distances are indeed
based on average distances. That is what the Poisson distribution
is
all about.

No, Sean, you are demonstrating your statistical ignorance again. The
Poisson distribution describes the probability of exactly k events
occuring
within a fixed interval,

That's right. How many targets are likely to be within a fixed
distance?

No Sean, read for comprehension. It is the probability of finding an
*exact
quantity* of events within the fixed interval. The probability mass
function
is measured over the *quantity* variable. It is not "how many targets are
likely to be found within a fixed distance."

The fixed interval is the fixed distance of X character changes. The
"events" are the finding of target sequenes within this fixed
distance. How likely are targets to be found within a particular
distance is most certainly a product of the the ratio and distribution
of targets and it most certainly does closely follow a Poisson
distribution.

No, Sean, that does not make sense. You used the word "ratio" without
defining what ratio you are taking, and a distribution is not a number - it
is a probability mass function. You don't seem to be comprehending what a
Poisson probability mass function does. But, as usual, your vague language
makes it impossible to know what you really mean.





given that the events occur at a known average rate

The average distance between targets is known.

lambda = n/interval, and that the occurance of next event is
independant
of
distance from previous event.

That's right . . .

Glad you agree, though above you clearly don't quite get it.

You have not demonstrated that you have met
the Poisson assumptions (especially event independance), and you have
incorrectly described what the Poisson distribution calculates.

The independence concept is only weakly relevant. The location of
targets in sequence space is not completely random. There is a loose
clustering effect. However, this effect is not significant enough to
affect the Poisson estimation to a significant degree.

Weaselly bull.

You need to look at BLAST data a bit more carefully.

Why don't you provide YOUR methodology for analyzing BLAST data in a new
thread? That should be interesting.



The only way there would be a significant effect is if Howard notion
of all the potentially beneficial targets being clustered in one tiny
corner of sequence/structure space was in fact correct. The actual
distribution is far more random in appearance than Howard and others
in this forum seem to realize.

And you know this how? From some misinterpreted abstract of an article
that
you didn't understand?

If you understand it any better, please do explain your take on what
the data actually says - i.e., like the data from the paper written by
Choi and Kim for starters.

The Choi and Kim paper did not say what you thought it did.

It's worth adding that the paper specifically states that it *can't*
show what he asserts that it shows.

RF

.

Relevant Pages

  • Re: Question about use of Poisson probabilities
    ... Probability theory was never my forte. ... I'm glad you're not going to call your boss an idiot. ... are a Poisson process, then 700,000 shows much, much, much too ... assumptions about your distribution that you need in order ...
    (sci.math)
  • Re: Planar Linkage Problem
    ... uniform distribution on [0,2pi) ... then of course the distance between endpoints is just 1. ... because there's zero probability that distance ...
    (sci.math)
  • Re: Poisson and exponential distributions
    ... i'm having some trouble seeing the relationship between the Poisson and the ... Because it is not a probability density, so does not need to integrate ... Thus, the complementary cumulative distribution of T is exp, ... if at least n arrivals occur by t. ...
    (sci.math)
  • Re: Probability question
    ... FreeCell in mind) and I want to calculate the probability that I win this ... Sum 0 to inf ) ... I intuitively suppose there is a relationship with a Poisson ... distribution but I don't see where p and n fit in, ...
    (sci.math)
  • Re: Why kullback-leibler distance?
    ... I am working with calculating the kullback-leibler distance between ... two probability distribution. ... standard to calculate the distances of PDF (probability ... symmetry. ...
    (sci.stat.edu)