Superposition. Thinning

Similar documents
Recall that by Theorems 10.3 and 10.4 together provide us the estimate o(n2 ), S(q) q 9, q=1

Where k is either given or determined from the data and c is an arbitrary constant.

The pn junction: 2 Current vs Voltage (IV) characteristics

Combinatorial Networks Week 1, March 11-12

Propositional Logic. Combinatorial Problem Solving (CPS) Albert Oliveras Enric Rodríguez-Carbonell. May 17, 2018

The Equitable Dominating Graph

10. The Discrete-Time Fourier Transform (DTFT)

6.1 Integration by Parts and Present Value. Copyright Cengage Learning. All rights reserved.

1 Minimum Cut Problem

The Matrix Exponential

cycle that does not cross any edges (including its own), then it has at least

The Matrix Exponential

Random Access Techniques: ALOHA (cont.)

Problem Statement. Definitions, Equations and Helpful Hints BEAUTIFUL HOMEWORK 6 ENGR 323 PROBLEM 3-79 WOOLSEY

Ch. 24 Molecular Reaction Dynamics 1. Collision Theory

Differential Equations

2. Laser physics - basics

Slide 1. Slide 2. Slide 3 DIGITAL SIGNAL PROCESSING CLASSIFICATION OF SIGNALS

Random Process Part 1

Chapter 10. The singular integral Introducing S(n) and J(n)

UNTYPED LAMBDA CALCULUS (II)

3. Poisson Processes (12/09/12, see Adult and Baby Ross)

Engineering 323 Beautiful HW #13 Page 1 of 6 Brown Problem 5-12

Manual for SOA Exam MLC.

Strongly Connected Components

SECTION where P (cos θ, sin θ) and Q(cos θ, sin θ) are polynomials in cos θ and sin θ, provided Q is never equal to zero.

SCHUR S THEOREM REU SUMMER 2005

y = 2xe x + x 2 e x at (0, 3). solution: Since y is implicitly related to x we have to use implicit differentiation: 3 6y = 0 y = 1 2 x ln(b) ln(b)

Chemical Physics II. More Stat. Thermo Kinetics Protein Folding...

Poisson Processes. Stochastic Processes. Feb UC3M

EEO 401 Digital Signal Processing Prof. Mark Fowler

Week 3: Connected Subgraphs

u 3 = u 3 (x 1, x 2, x 3 )

Introduction to Arithmetic Geometry Fall 2013 Lecture #20 11/14/2013

Equidistribution and Weyl s criterion

Section 6.1. Question: 2. Let H be a subgroup of a group G. Then H operates on G by left multiplication. Describe the orbits for this operation.

NEW APPLICATIONS OF THE ABEL-LIOUVILLE FORMULA

Functions of Two Random Variables

ON RIGHT(LEFT) DUO PO-SEMIGROUPS. S. K. Lee and K. Y. Park

10. EXTENDING TRACTABILITY

Solution of Assignment #2

Lecture 37 (Schrödinger Equation) Physics Spring 2018 Douglas Fields

Basic Polyhedral theory

orbiting electron turns out to be wrong even though it Unfortunately, the classical visualization of the

a 1and x is any real number.

MSLC Math 151 WI09 Exam 2 Review Solutions

COHORT MBA. Exponential function. MATH review (part2) by Lucian Mitroiu. The LOG and EXP functions. Properties: e e. lim.

Supplementary Materials

Addition of angular momentum

Einstein Equations for Tetrad Fields

Functions of Two Random Variables

Computing and Communications -- Network Coding

2.3 Matrix Formulation

Southern Taiwan University

Poisson Processes. Particles arriving over time at a particle detector. Several ways to describe most common model.

Renewal theory and its applications

MEMORIAL UNIVERSITY OF NEWFOUNDLAND

Exponential Distribution and Poisson Process

Radiation Physics Laboratory - Complementary Exercise Set MeBiom 2016/2017

Continuous probability distributions

Thus, because if either [G : H] or [H : K] is infinite, then [G : K] is infinite, then [G : K] = [G : H][H : K] for all infinite cases.

ANALYSIS IN THE FREQUENCY DOMAIN

Probability and Stochastic Processes: A Friendly Introduction for Electrical and Computer Engineers Roy D. Yates and David J.

Observer Bias and Reliability By Xunchi Pu

Brief Introduction to Statistical Mechanics

BINOMIAL COEFFICIENTS INVOLVING INFINITE POWERS OF PRIMES

First derivative analysis

Hydrogen Atom and One Electron Ions

Some remarks on Kurepa s left factorial

ECE602 Exam 1 April 5, You must show ALL of your work for full credit.

Bayesian Decision Theory

Addition of angular momentum

Cramér-Rao Inequality: Let f(x; θ) be a probability density function with continuous parameter

COMPUTER GENERATED HOLOGRAMS Optical Sciences 627 W.J. Dallas (Monday, April 04, 2005, 8:35 AM) PART I: CHAPTER TWO COMB MATH.

SOME PARAMETERS ON EQUITABLE COLORING OF PRISM AND CIRCULANT GRAPH.

Need to understand interaction of macroscopic measures

MA 262, Spring 2018, Final exam Version 01 (Green)

(Upside-Down o Direct Rotation) β - Numbers

On the irreducibility of some polynomials in two variables

That is, we start with a general matrix: And end with a simpler matrix:

Construction of asymmetric orthogonal arrays of strength three via a replacement method

Objective Mathematics

Math 34A. Final Review

EXST Regression Techniques Page 1

Abstract Interpretation: concrete and abstract semantics

Square of Hamilton cycle in a random graph

BINOMIAL COEFFICIENTS INVOLVING INFINITE POWERS OF PRIMES. 1. Statement of results

Search sequence databases 3 10/25/2016

Higher order derivatives

CPSC 665 : An Algorithmist s Toolkit Lecture 4 : 21 Jan Linear Programming

Bifurcation Theory. , a stationary point, depends on the value of α. At certain values

Alpha and beta decay equation practice

Linear-Phase FIR Transfer Functions. Functions. Functions. Functions. Functions. Functions. Let

1 Input-Output Stability

4 x 4, and. where x is Town Square

Function Spaces. a x 3. (Letting x = 1 =)) a(0) + b + c (1) = 0. Row reducing the matrix. b 1. e 4 3. e 9. >: (x = 1 =)) a(0) + b + c (1) = 0

u x v x dx u x v x v x u x dx d u x v x u x v x dx u x v x dx Integration by Parts Formula

Chapter 1. Chapter 10. Chapter 2. Chapter 11. Chapter 3. Chapter 12. Chapter 4. Chapter 13. Chapter 5. Chapter 14. Chapter 6. Chapter 7.

Lecture 2: Discrete-Time Signals & Systems. Reza Mohammadkhani, Digital Signal Processing, 2015 University of Kurdistan eng.uok.ac.

BSc Engineering Sciences A. Y. 2017/18 Written exam of the course Mathematical Analysis 2 August 30, x n, ) n 2

Transcription:

Suprposition STAT253/317 Wintr 213 Lctur 11 Yibi Huang Fbruary 1, 213 5.3 Th Poisson Procsss 5.4 Gnralizations of th Poisson Procsss Th sum of two indpndnt Poisson procsss with rspctiv rats λ 1 and λ 2, calld th suprposition of th procsss, is again a Poisson procss but with rat λ 1 + λ 2. Th proof is straight forward from Dfinition 5.3 and hnc omittd. Rmark: By rpatd application of th abov argumnts w can s that th suprposition of k indpndnt Poisson procsss with rats λ 1,, λ n is again a Poisson procss with rat λ 1 + + λ n. STAT253/317 213 Wintr Lctur 11-1 STAT253/317 213 Wintr Lctur 11-2 Why Poisson Procsss Mak Sns? Thr is a usful rsult in probability thory which says that: Hr if w tak N indpndnt counting procsss and sum thm up, thn th rsulting suprposition procss is approximatly a Poisson procss. N must b larg nough and th rats of th individual procsss must b small rlativ to N but th individual procsss that go into th suprposition can othrwis b arbitrary. Thinning Considr a Poisson procss with rat λ. At ach arrival of vnts, it is classifid as a { Typ 1 vnt with probability p or Typ 2 vnt with probability 1 p, indpndntly of all othr vnts. Lt N i (t) # of typ i vnts occurrd during [, t], i 1, 2. Not that N(t) N 1 (t) + N 2 (t). Proposition 5.2 {N 1 (t), t } and {N 2 (t), t } ar both Poisson procsss having rspctiv rats λp and λ(1 p). Furthrmor, th two procsss ar indpndnt STAT253/317 213 Wintr Lctur 11-3 STAT253/317 213 Wintr Lctur 11-4

Proof of Proposition 5.2 First obsrv that givn N(t) n + m, (N 1 (t), N 2 (t)) Binomial(n + m, p, (1 p)). (why?) Thus P(N 1 (t) n, N 2 (t) m) P(N 1 (t) n, N 2 (t) m N(t) n + m)p(n(t) n + m) ( n + m )p n (1 p) m λt (λt)n+m n (n + m)! λtp (λpt)n λt(1 p) (λ(1 p)t)m n! m! P(N 1 (t) n)p(n 2 (t) m). This provs th indpndnc of N 1 (t) and N 2 (t) and that N 1 (t) Poisson(λpt), N 2 (t) Poisson(λ(1 p)t). Both {N 1 (t)} and {N 2 (t)} inhrit th stationary and indpndnt incrmnt proprtis from {N(t)}, and hnc ar both Poisson procsss. STAT253/317 213 Wintr Lctur 11-5 Som Convrs of Thinning & Suprposition (Cont d) Lt N(t) N A (t) + N B (t) b th suprposition of th two procsss. Lt I i { 1 if th ith vnt in th suprpositon procss is an A vnt othrwis Th I i, i 1, 2,... ar i.i.d. Brnoulli(p). whr p Approach 2: λ A λ A + λ B. P(S A n < S B 1 ) P(th first n vnts ar all A vnts) ( λa λ A + λ B P(S A n < S B m) P(at last n A vnts occur bfor m B vnts) ) n P(gtting at last n hads bfor m tails) n+m 1 ( ) ( ) n + m 1 k ( λa λb k λ A + λ B λ A + λ B kn. ) n+m 1+k STAT253/317 213 Wintr Lctur 11-7 Som Convrs of Thinning & Suprposition Considr two indp. Poisson procsss {N A (t)} and {N B (t)} w/ rspctiv rats λ A and λ B. Lt Find P(S A n < S B m). S A n arrival tim of th mth A vnt S B m arrival tim of th nth B vnt Approach 1: Obsrvr that S A n Gamma(m, λ A ), S B m Gamma(n, λ B ) and thy ar indpndnt. Thus P(Sn A < Sm) B x<y λ A λ Ax (λ Ax) n 1 (n 1)! λ B λ By (λ By) m 1 (m 1)! dxdy STAT253/317 213 Wintr Lctur 11-6 Proposition 5.3 (Gnralization of Prop. 5.2) Considr a Poisson procss with rat λ. If an vnt occurs at tim t will b classifid as a typ i vnt with probability p i (t), i 1,..., k, i p i(t) 1, for all t, indpndntly of all othr vnts. thn N i (t) numbr of typ i vnts occurring in [, t], i 1,..., k. Not N(t) k N i(t). Thn N i (t), i 1,..., k ar indpndnt Poisson random variabls with mans λ p i(s)ps STAT253/317 213 Wintr Lctur 11-8

Exampl (Rvision of Exrcis 6.66 on p.364) Policyholdrs of a crtain insuranc company hav accidnts occurring according to a Poisson procss with rat λ. Th amount of tim T from whn th accidnt occurs until a claim is mad has distribution G(t) P(T t). Lt N c (t) b th numbr of claims mad by tim t. Find th distribution of N c (t). Solution. Considr an accidnt occurrd at tim s. Thn it is claimd by tim t if s + T t, i.., with probability p(s) P(T t s) G(t s). By Proposition 5.3, N c (t) has a Poisson distribution with man λ p(s)ps λ G(t s)ds λ G(s)ds STAT253/317 213 Wintr Lctur 11-9 Proposition 5.4 Lt {N 1 (t), t }, and {N 2 (t), t } b two indpndnt nonhomognous Poisson procss with rspctiv intnsity functions λ 1 (t) and λ 2 (t), and lt N(t) N 1 (t) + N 2 (t). Thn (a) {N(t), t } is a nonhomognous Poisson procss with intnsity function λ 1 (t) + λ 2 (t). (b) Givn that an vnt of th {N(t), t } procss occurs at tim t thn, indpndnt of what occurrd prior to t, th vnt at t was from th {N 1 (t)} procss with probability λ 1 (t) λ 1 (t) + λ 2 (t). STAT253/317 213 Wintr Lctur 11-11 5.4.1 Nonhomognous Poisson Procss Dfinition 5.4a. A nonhomognous (a.k.a. non-stationary) Poisson procss with intnsity function λ(t) is a counting procss {N(t), t } satisfying (i) N(). (ii) having indpndnt incrmnts. (iii) P(N(t + h) N(t) 1) λ(t)h + o(h). (iv) P(N(t + h) N(t) 2) o(h). Dfinition 5.4b. A nonhomognous Poisson procss with intnsity function λ(t) is a counting procss {N(t), t } satisfying (i) N(), (ii) for s, t, N(t + s) N(s) is indpndnt of N(s) (indpndnt incrmnt) (iii) For s, t, N(t + s) N(s) Poisson(m(t + s) m(s)), whr m(t) λ(t)dt Th two dfinitions ar quivalnt STAT253/317 213 Wintr Lctur 11-1 5.4.2 Compound Poisson Procsss Dfinition. Lt {N(t)} b a (homognous) Poisson procss with rat λ and Y 1, Y 2,... ar i.i.d random variabls indpndnt of {N(t)}. Th procss X (t) N(t) Y i is calld a compound Poisson procss, in which X (t) is dfind as if N(t). A compound Poisson procss has indpndnt incrmnt, sinc X (t + s) X (s) N(t+s) N(s) Y i+n(s) is indpndnt of X (s) N(s) Y i, and stationary incrmnt, sinc X (t + s) X (s) N(t+s) N(s) Y i+n(s) has th sam distribution as X (t) N(t) Y i STAT253/317 213 Wintr Lctur 11-12

Th Man of a Compound Poisson Procss Suppos E[Y i ] µ Y, Var(Y i ) σy 2. Not that E[N(t)] λt. E[X (t) N(t)] N(t) E[Y i N(t)] N(t) E[Y i] (sinc Y i s ar indp. of N(t)) N(t)µ Y Thus E[X (t)] E[E[X (t) N(t)]] E[N(t)]µ Y λtµ Y Varianc of a Compound Poisson Procss (Cont d) Similarly, using that E[N(t)] Var(N(t)) λt, w hav ( N(t) ) Var[X (t) N(t)] Var Y i N(t) N(t) Var(Y i N(t)) N(t) Var[Y i] (sinc Y i s ar indp. of N(t)) N(t)σ 2 Y E[V [X (t) N(t)]] E[N(t)]σ 2 Y λtσ2 Y Var(E[X (t) N(t)]) Var(N(t)µ Y ) Var(N(t))µ 2 Y λtµ2 Y Thus Var(X (t)) E[V [X (t) N(t)]] + Var(E[X (t) N(t)]) λt(σ 2 Y + µ2 Y ) λte[y 2 i ] STAT253/317 213 Wintr Lctur 11-13 STAT253/317 213 Wintr Lctur 11-14 CLT of a Compound Poisson Procss 5.6.3 Conditional Poisson Procsss As t, th distribution of X (t) E[X (t)] X (t) λtµ Y Var(X (t)) λt(σy 2 + µ2 Y ) convrgs to a standard normal distribution N(, 1). Dfinition. A conditional (or mixd) Poisson procss {N(t), t } is a counting procss satisfying (i) N(), (ii) having stationary incrmnt, and (iii) thr is a random variabl Λ > with probability dnsity g(λ), such that N(t + s) N(s) Λ Poisson(λt), i.., P(N(t + s) N(s) k) g(λ)dλ, k, 1,... STAT253/317 213 Wintr Lctur 11-15 STAT253/317 213 Wintr Lctur 11-16

Rmark: In gnral, a conditional Poisson procss dos NOT hav indpndnt incrmnt. P(N(s) j, N(t + s) N(s) k) ( λs (λs)j j! λs (λs)j g(λ)dλ j! g(λ)dλ ) ( P(N(s) j)p(n(t + s) N(s) k) ) g(λ)dλ STAT253/317 213 Wintr Lctur 11-17