Random measure explained

In probability theory, a random measure is a measure-valued random element. Random measures are for example used in the theory of random processes, where they form many important point processes such as Poisson point processes and Cox processes.

Definition

Random measures can be defined as transition kernels or as random elements. Both definitions are equivalent. For the definitions, let

E

be a separable complete metric space and let

lE

be its Borel

\sigma

-algebra
. (The most common example of a separable complete metric space is

\Rn

)

As a transition kernel

A random measure

\zeta

is a (a.s.) locally finite transition kernel from an abstract probability space

(\Omega,lA,P)

to

(E,lE)

.

Being a transition kernel means that

B\inllE

, the mapping

\omega\mapsto\zeta(\omega,B)

is measurable from

(\Omega,lA)

to

(\R,lB(\R))

\omega\in\Omega

, the mapping

B\mapsto\zeta(\omega,B)(B\inlE)

is a measure on

(E,lE)

Being locally finite means that the measures

B\mapsto\zeta(\omega,B)

satisfy

\zeta(\omega,\tildeB)<infty

for all bounded measurable sets

\tildeB\inlE

and for all

\omega\in\Omega

except some

P

-null set

In the context of stochastic processes there is the related concept of a stochastic kernel, probability kernel, Markov kernel.

As a random element

Define

\tildelM:=\{\mu\mid\muismeasureon(E,lE)\}

and the subset of locally finite measures by

lM:=\{\mu\in\tildelM\mid\mu(\tildeB)<inftyforallboundedmeasurable\tildeB\inlE\}

For all bounded measurable

\tildeB

, define the mappings

I\tilde\colon\mu\mapsto\mu(\tildeB)

from

\tildelM

to

\R

. Let

\tildeM

be the

\sigma

-algebra induced by the mappings

I\tilde

on

\tildelM

and

M

the

\sigma

-algebra induced by the mappings

I\tilde

on

lM

. Note that

\tildeM|lM=M

.

A random measure is a random element from

(\Omega,lA,P)

to

(\tildelM,\tildeM)

that almost surely takes values in

(lM,M)

Basic related concepts

Intensity measure

See main article: article and intensity measure. For a random measure

\zeta

, the measure

\operatornameE\zeta

satisfying

\operatornameE\left[\intf(x)\zeta(dx)\right]=\intf(x)\operatornameE\zeta(dx)

for every positive measurable function

f

is called the intensity measure of

\zeta

. The intensity measure exists for every random measure and is a s-finite measure.

Supporting measure

For a random measure

\zeta

, the measure

\nu

satisfying

\intf(x)\zeta(dx)=0a.s.iff\intf(x)\nu(dx)=0

for all positive measurable functions is called the supporting measure of

\zeta

. The supporting measure exists for all random measures and can be chosen to be finite.

Laplace transform

For a random measure

\zeta

, the Laplace transform is defined as

lL\zeta(f)=\operatornameE\left[\exp\left(-\intf(x)\zeta(dx)\right)\right]

for every positive measurable function

f

.

Basic properties

Measurability of integrals

For a random measure

\zeta

, the integrals

\intf(x)\zeta(dx)

and

\zeta(A):=\int1A(x)\zeta(dx)

for positive

lE

-measurable

f

are measurable, so they are random variables.

Uniqueness

The distribution of a random measure is uniquely determined by the distributions of

\intf(x)\zeta(dx)

for all continuous functions with compact support

f

on

E

. For a fixed semiring

lI\subsetlE

that generates

lE

in the sense that

\sigma(lI)=lE

, the distribution of a random measure is also uniquely determined by the integral over all positive simple

lI

-measurable functions

f

.

Decomposition

A measure generally might be decomposed as:

\mu=\mud+\mua=\mud+

N
\sum
n=1

\kappan

\delta
Xn

,

Here

\mud

is a diffuse measure without atoms, while

\mua

is a purely atomic measure.

Random counting measure

A random measure of the form:

N
\mu=\sum
n=1
\delta
Xn

,

where

\delta

is the Dirac measure, and

Xn

are random variables, is called a point process or random counting measure. This random measure describes the set of N particles, whose locations are given by the (generally vector valued) random variables

Xn

. The diffuse component

\mud

is null for a counting measure.

In the formal notation of above a random counting measure is a map from a probability space to the measurable space a measurable space. Here

NX

is the space of all boundedly finite integer-valued measures

N\inMX

(called counting measures).

The definitions of expectation measure, Laplace functional, moment measures and stationarity for random measures follow those of point processes. Random measures are useful in the description and analysis of Monte Carlo methods, such as Monte Carlo numerical quadrature and particle filters.

See also

References

[1] [2] [3] [4] [5] [6] [7]

Notes and References

  1. Book: Klenke . Achim . 2008 . Probability Theory . Berlin . Springer . 10.1007/978-1-84800-048-3 . 978-1-84800-047-6 . 526.
  2. Book: Kallenberg . Olav . Olav Kallenberg . 2017 . Random Measures, Theory and Applications. Probability Theory and Stochastic Modelling . 77 . Switzerland . Springer . 1. 10.1007/978-3-319-41598-7. 978-3-319-41596-3.
  3. "Crisan, D., Particle Filters: A Theoretical Perspective, in Sequential Monte Carlo in Practice, Doucet, A., de Freitas, N. and Gordon, N. (Eds), Springer, 2001,
  4. Book: 10.1007/b97277 . D. J. . Daley . D. . Vere-Jones. An Introduction to the Theory of Point Processes . Probability and its Applications . 2003 . 0-387-95541-0 .
  5. [Olav Kallenberg|Kallenberg, O.]
  6. Jan Grandell, Point processes and random measures, Advances in Applied Probability 9 (1977) 502-526. JSTOR A nice and clear introduction.
  7. Book: Kallenberg . Olav . Olav Kallenberg . 2017 . Random Measures, Theory and Applications. Probability Theory and Stochastic Modelling . 77 . Switzerland . Springer . 52. 10.1007/978-3-319-41598-7. 978-3-319-41596-3.