Science > Physics > Logistic Distribution Comes From Rarity in arXiv and out of arXiv
| Topic: |
Science > Physics |
| User: |
"OsherD" |
| Date: |
30 Sep 2005 10:34:45 PM |
| Object: |
Logistic Distribution Comes From Rarity in arXiv and out of arXiv |
From Osher Doctorow
The Riccati Differential equation is the fundamental equation of
expansion-contraction in PI (see my earlier threads), and the Logistic
Differential equation is a subcase of the Riccati, which also can be
generalized to the Logistic probability distribution.
Now Michael Baake (U. Bielefeld, Germany), Uwe Grimm (Open U., U.K.),
and Harald Jockusch (U. Bielefeld) in "Trying to be rare,"
q-bio.P.W./0509011 v2 14 Sep 2005 in Front For the Mathematics arXiv (7
pages) prove that under rather general conditions, for a large but
finite population of N individuals with N constant divided into M grops
and if there is an advantage in belonging to a smaller group but
changing the group only occurs if individuals from different groups
meet, then in the infinite population limit the dynamical system has an
explicit solution built piecewise from solutions of the Logistic
differential equation in continuous time.
Baake et al haven't communicated with me so presumably don't know
Probable Influence/Causation (PI), and I have some differences with
their interpretation of the above as a game theoretic scenario with use
of nonlocal information (the group sizes). They don't actually use
game theory in their paper, but it is still of some importance to
clarify whether group size is local or nonlocal or both. We're not
talking about group velocity here but the number of Individuals in a
Plurality (group). The Plurality here is measured in number as nothing
but the number of Individuals. While Number here extends beyond one
Individual, it never extends beyond "any" one individual on each
increment of the counter.
The confusion between Individual and Plurality arises for many people
in various interdisciplinary contexts including politics because we all
are brought up when young to distinguish between our internal world
("Me First, Number One") and our external world (everybody who's not
"Me"). But objectively there are more than one Individual in the real
(physical) world, so Locality cannot be the claim that there is only
one Individual ("Me"). Locality is the claim that nothing happens
among Pluralities as well as Individuals which is not directly carried
out or enacted or instantiated by Individuals. Nonlocality would exist
if there were some "group consciousness" independent of Individuals,
"racial consciousness" independent of individuals, or some
chemical-like scenario for example which although starting out with
Individual elements is a creation of something beyond Individuals and
not enacted by Individuals.
We can be "kind-hearted" to the Nonlocalists and refer to "almost
Nonlocal" or even "Nonlocal" if we understand the qualification that
we're referring to something that happens outside one particular fixed
Individual. We could even benefit somewhat by comparing a particular
fixed Individual with everybody or many Individuals outside him/her.
But the most fundamental difference is not captured by this
"kind-heartedness".
Osher Doctorow
.
|
|
| User: "OsherD" |
|
| Title: Re: Logistic Distribution Comes From Rarity in arXiv and out of arXiv |
30 Sep 2005 10:42:49 PM |
|
|
From Osher Doctorow
Baake et al refer to examples of nonlocal information applications as
spatial advantage (population patches), supply advantage (food
specialization), promotional advantage (advertising strategies),
recognizability advantages (name choosing), etc. Yet it would be
absurd to regard supply as nonlocal, advertising as nonlocal, name
choosing as nonlocal, or population patches such as a large population
and its region as nonlocal. If an Individual is doing it, it's local.
It doesn't have to be the same Inidividual at every spatial or temporal
location of the action or process or thing or set, but it has to be
done by an Individual at each time and place.
Osher Doctorow
.
|
|
|
| User: "OsherD" |
|
| Title: Re: Logistic Distribution Comes From Rarity in arXiv and out of arXiv |
30 Sep 2005 10:48:24 PM |
|
|
From Osher Doctorow
The citation should be q-bio.PE/0509011, not q-bio.P.W./0509011.
Take a look also at Antar Bandyopadhyay's two papers, math.PR/0407175
and 0401389, both 2004 (10 Jul vs 28 Jan), "Bivariate uniqueness and
endogeny for recursive distributional equations: two examples," and
"Bivariate uniqueness in the logistic recursive distributional
equation," respectively.
Osher Doctorow
.
|
|
|
| User: "OsherD" |
|
| Title: Re: Logistic Distribution Comes From Rarity in arXiv and out of arXiv |
30 Sep 2005 10:53:52 PM |
|
|
From Osher Doctorow
Bandyopadhyay is at Institute for Mathematics and its Applications, U.
Minnesota.
Osher Doctorow
.
|
|
|
| User: "OsherD" |
|
| Title: Re: Logistic Distribution Comes From Rarity in arXiv and out of arXiv |
30 Sep 2005 11:30:54 PM |
|
|
From Osher Doctorow
Bandyopadhyay's 10 Jul 2004 paper cited has what looks at first glance
like a super-abstract definition of "endogeny", namely whether the
associated RTP (recursive tree process, a tree-indexed random process
whose variables are often used as auxiliary variables to define some
useful random structure) is measurable with respect to the data, namey
independent identically distributed (iid) copies of the pair (z, N)
where a general distributional identity looks like:
1) X = -d g(z; (Xj: 1 < = j < = N))
where the (Xj) are iid with same distribution as X and independent of
pair (z, N) with N positive integer-valued random variable, g a given
function, and = -d means "has the same distribution as" and I've left
out one technicality involving the last inequality < = N for
simplicity.
While it is interesting to know that several quite different looking
problems (one of which is Logistic) both yield exogeny (which may have
been Bandyopadhyay's motivation), in some ways it is even more
interesting to notice that other things relate the different looking
problem which Bandyopadhyay gives as examples. The first example is
the frozen percolation process on an infinite regular binary tree,
where z is the uniform[0, 1] distributed random variable which by now
we hopefully realize is the best PI-optimal distribution. The second
example arises from study of the asymptotic limit of random assignment
problem with the RDE (recursive distributional equations) given by:
2) X = -d min(zj - Xj) on R, (Xj) iid with same law/distribution as X
and independent of (zj) which are points of a Poisson point process of
rate 1 on (0, infinity)
which is called the Logistic RDE because its unique solution is the
Logistic distribution:
3) P(X < = x) =1/(1 + exp(-x)), x real
Equation (2) is especially interesting because PI involves differences
1 + y - x where y = P(AB), x = P(A), or for the second version of PI y
= P(B), y < = x, but this generalizes to any differences as previous
postings have elaborated, and contrasts with division methods such as
conditional probability. The more general distinction is between
addition/subtraction on the one hand and multiplication/division on the
other hand, although from the viewpoint of PI it is arguably
subtraction and division which "lead the way" in each subtype.
So we are finding several cross-research results which agree that
Probable Influence (PI) or its generalizations are associated with Rare
Events, Subtraction, Riccati (including Logistic) Differential
equations, in contrast with conditional probability or its
generalizations which is not associated with Rare Events but is
associated with Division and with Markov type and other Non-Riccati
Differential equations including Fokker-Planck and Chapman-Kolmogorov.
Osher Doctorow
.
|
|
|
|
|
|
|
|
Related Articles |
|
|
