Lecture 6. Poisson processes

Similar documents
Lecture 8. Confidence intervals and the central limit theorem

Important Probability Distributions OPRE 6301

Example: 1. You have observed that the number of hits to your web site follow a Poisson distribution at a rate of 2 per day.

MAS108 Probability I

Notes on Continuous Random Variables

LECTURE 16. Readings: Section 5.1. Lecture outline. Random processes Definition of the Bernoulli process Basic properties of the Bernoulli process

2. Discrete random variables

Overview of Monte Carlo Simulation, Probability Review and Introduction to Matlab

ECE302 Spring 2006 HW3 Solutions February 2,

Introduction to Probability

5. Continuous Random Variables

Homework set 4 - Solutions

How To Find Out How Much Money You Get From A Car Insurance Claim

Lecture Notes 1. Brief Review of Basic Probability

Stochastic Processes and Queueing Theory used in Cloud Computer Performance Simulations

ECE302 Spring 2006 HW4 Solutions February 6,

Definition: Suppose that two random variables, either continuous or discrete, X and Y have joint density

Exponential Distribution

Joint Exam 1/P Sample Exam 1

Poisson process. Etienne Pardoux. Aix Marseille Université. Etienne Pardoux (AMU) CIMPA, Ziguinchor 1 / 8

Tenth Problem Assignment

e.g. arrival of a customer to a service station or breakdown of a component in some system.

Math 431 An Introduction to Probability. Final Exam Solutions

Poisson Processes. Chapter Exponential Distribution. The gamma function is defined by. Γ(α) = t α 1 e t dt, α > 0.

The Exponential Distribution

6.041/6.431 Spring 2008 Quiz 2 Wednesday, April 16, 7:30-9:30 PM. SOLUTIONS

Random variables, probability distributions, binomial random variable

IEOR 6711: Stochastic Models, I Fall 2012, Professor Whitt, Final Exam SOLUTIONS

What is Statistics? Lecture 1. Introduction and probability review. Idea of parametric inference

Lecture 5 : The Poisson Distribution

CHAPTER 6: Continuous Uniform Distribution: 6.1. Definition: The density function of the continuous random variable X on the interval [A, B] is.

FEGYVERNEKI SÁNDOR, PROBABILITY THEORY AND MATHEmATICAL

Random variables P(X = 3) = P(X = 3) = 1 8, P(X = 1) = P(X = 1) = 3 8.

Sums of Independent Random Variables

Math 370/408, Spring 2008 Prof. A.J. Hildebrand. Actuarial Exam Practice Problem Set 2 Solutions

Statistics 100A Homework 4 Solutions

M/M/1 and M/M/m Queueing Systems

2 Binomial, Poisson, Normal Distribution

6.2. Discrete Probability Distributions

Stat 515 Midterm Examination II April 6, 2010 (9:30 a.m. - 10:45 a.m.)

Chapter 4 Lecture Notes

Statistics 100A Homework 7 Solutions

Modelling Patient Flow in an Emergency Department

IEOR 6711: Stochastic Models I Fall 2012, Professor Whitt, Tuesday, September 11 Normal Approximations and the Central Limit Theorem

Wald s Identity. by Jeffery Hein. Dartmouth College, Math 100

Section 5.1 Continuous Random Variables: Introduction

Feb 7 Homework Solutions Math 151, Winter Chapter 4 Problems (pages )

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 5 9/17/2008 RANDOM VARIABLES

Chapter 4. Probability Distributions

CHAPTER 7 SECTION 5: RANDOM VARIABLES AND DISCRETE PROBABILITY DISTRIBUTIONS

Lecture 7: Continuous Random Variables

Jan Vecer Tomoyuki Ichiba Mladen Laudanovic. Harvard University, September 29, 2007

ANALYZING NETWORK TRAFFIC FOR MALICIOUS ACTIVITY

uation of a Poisson Process for Emergency Care

Math 461 Fall 2006 Test 2 Solutions

STAT 315: HOW TO CHOOSE A DISTRIBUTION FOR A RANDOM VARIABLE

Practice Problems #4

Week 2: Exponential Functions

Marshall-Olkin distributions and portfolio credit risk

Poisson processes (and mixture distributions)

Some special discrete probability distributions

Chapter 5 Discrete Probability Distribution. Learning objectives

Chapter 5. Random variables

STT315 Chapter 4 Random Variables & Probability Distributions KM. Chapter 4.5, 6, 8 Probability Distributions for Continuous Random Variables

Homework 4 - KEY. Jeff Brenion. June 16, Note: Many problems can be solved in more than one way; we present only a single solution here.

6.263/16.37: Lectures 5 & 6 Introduction to Queueing Theory

), 35% use extra unleaded gas ( A

Probability density function : An arbitrary continuous random variable X is similarly described by its probability density function f x = f X

Queueing Systems. Ivo Adan and Jacques Resing

LECTURE - 1 INTRODUCTION TO QUEUING SYSTEM

Ch5: Discrete Probability Distributions Section 5-1: Probability Distribution

Military Reliability Modeling William P. Fox, Steven B. Horton

Discrete Mathematics and Probability Theory Fall 2009 Satish Rao, David Tse Note 18. A Brief Introduction to Continuous Probability

Master s Theory Exam Spring 2006

The Normal Distribution. Alan T. Arnholt Department of Mathematical Sciences Appalachian State University

Probability Generating Functions

How to Gamble If You Must

STAT x 0 < x < 1

10 GEOMETRIC DISTRIBUTION EXAMPLES:

8.7 Exponential Growth and Decay

Chapter 3: DISCRETE RANDOM VARIABLES AND PROBABILITY DISTRIBUTIONS. Part 3: Discrete Uniform Distribution Binomial Distribution

PROBABILITY AND SAMPLING DISTRIBUTIONS

2WB05 Simulation Lecture 8: Generating random variables

Random Variables. Chapter 2. Random Variables 1

Question: What is the probability that a five-card poker hand contains a flush, that is, five cards of the same suit?

Gambling and Data Compression

6 POISSON DISTRIBUTIONS

Solutions to Problem Set 1

UNIT 2 QUEUING THEORY

Discrete Mathematics and Probability Theory Fall 2009 Satish Rao, David Tse Note 13. Random Variables: Distribution and Expectation

Chapter 4 - Lecture 1 Probability Density Functions and Cumul. Distribution Functions

ECON1003: Analysis of Economic Data Fall 2003 Answers to Quiz #2 11:40a.m. 12:25p.m. (45 minutes) Tuesday, October 28, 2003

Chapter 9 Monté Carlo Simulation

TEST 2 STUDY GUIDE. 1. Consider the data shown below.

CHAPTER 3 CALL CENTER QUEUING MODEL WITH LOGNORMAL SERVICE TIME DISTRIBUTION

November 2000 Course 3

. (3.3) n Note that supremum (3.2) must occur at one of the observed values x i or to the left of x i.

Basic Queueing Theory

The Binomial Probability Distribution

Characteristics of Binomial Distributions

Transcription:

Lecture 6. Poisson processes Mathematical Statistics and Discrete Mathematics November 18th, 2015 1 / 15

Motivation We want to have a probabilistic model describing the number of calls arriving at a call center within a given time range, say, 15 minutes between 1pm and 1.15pm. We can represent the incoming calls as marks on the timeline: 0 5 10 15 So that it is a realistic model, it should posses the following properties: The numbers of calls arriving in disjoint time intervals should be independent. 0 5 10 15 The distribution of the number of calls arriving in a certain time interval depends only on the length of the interval and not its placement in time. 0 5 10 15 We will define a stochastic process which satisfies these properties, namely the Poisson process. 2 / 15

Poisson distribution A discrete random variable X has Poisson distribution with parameter λ > 0 if f X (k) = P(X = k) = λk k! e λ, for k = 0, 1, 2,... We denote this by writing X Pois(λ). Properties: E[X] = Var[X] = λ. If X Pois(λ 1 ), Y Pois(λ 2 ), and X and Y are independent, then X + Y Pois(λ 1 + λ 2 ). Pois(λ) can be well approximated by Binom(n, p) with very large n and small p such that np = λ. Hence, it models well the number of successes in a large number of trials where the success probability is small. This is sometimes called the law of rare events. 3 / 15

Poisson distribution Poisson distribution can be used to model the number of people winning the lottery each week, people older than 100 years in some community, deaths in a given age group within a year, mutations in a piece of DNA after exposure to radiation, times a server is accessed per minute, telephone calls arriving at a call center within 15 minutes (the population covered by one call center is large and each person has a small probability of making a call within a particular 15 minutes period). 4 / 15

Exponential distribution A continous random variable X has exponential distribution with parameter λ > 0 if We denote this by writing X Exp(λ). f X (x) = λe λx for x > 0. Properties: E[X] = λ 1, Var[X] = λ 2. F X (x) = P(X x) = 1 e λx for x > 0. 5 / 15

Exponential distribution A continuous random variable X has exponential distribution, if and only if it is memoryless, that is P(X > s + t X > t) = P(X > s), for any s, t > 0. Proof. Suppose X has exponential distribution with parameter λ. Then, we have for any x > 0, and hence P(X > x) = 1 P(X x) = e λx, P(X > s + t) = P(X > s) P(X > t) by the property of the exponential function. The opposite implication follows since the exponential function is the only one satisfying this property, and hence the CDF function is uniquely identified. The only discrete memoryless distribution is the geometric distribution. 6 / 15

Exponential distribution Let X be the time until the arrival of the first call at the call center. Since we assume that the arrivals are independent in disjoint intervals, the knowledge that there was no call until time s (this is the event {X > t}, has no influence on the arrivals after time s). Moreover, since the arrivals after time s in an interval of length t should be distributed as the arrivals in the interval [0, t], we get P(X > s + t X > t) = P(X > s), for any s, t > 0, and hence X has exponential distribution. Time until a radioactive particle decays can be modelled with exponential distribution. Time between server requests can be modelled with exponential distribution. 7 / 15

Counting process A counting process is stochastic process {C t } t 0 that keeps count of events that have occurred up to time t. In particular, C t it takes values in {0, 1, 2,...}, C t it is non-decreasing in t, C t C s is the number of events in the interval (s, t] for s < t. Here, the word event has the traditional, and not probabilistic, meaning. Counting processes are also called jump processes. 8 / 15

Counting process The graph below shows an instance of a counting process together with the times when the events occurred: 30 25 20 15 10 5 2 4 6 8 10 12 14 0 5 10 15 9 / 15

Poisson process A Poisson process {N t } t 0 with intensity parameter λ > 0 is a counting process with the following additional properties: (starting at 0) N 0 = 0. (independent increments) The random variables (N t2 N t1 ) and (N t4 N t3 ) are independent for all t 1 < t 2 < t 3 < t 4. (stationarity) Fix r > 0. The random variable (N t+r N t ) has the same distribution for all t. (no double jumps/events) P(N h = 1) h λ, and P(N h 2) h 0, as h 0. The last property says that the probability that two events counted by the Poisson process happen at the same time is 0. 10 / 15

Poisson process Calls arriving at a call center can be modelled by a Poisson process. Request arriving at a server can be modelled by a Poisson process. Arrivals of students for this lecture cannot be modelled by a Poisson process. 11 / 15

Properties of Poisson processes If {N t } t 0 is a Poisson process with intensity λ, then: N t Pois(λt) for all t. In particular, E[N t ] = λt. N t N s Pois(λ(t s)) for all s < t. In particular, E[N t N s ] = λ(t s). Let T 0 = 0, T 1, T 2, T 3,... be the (random) times when the events counted by the process occur. Let τ i = T i T i 1, for i = 1, 2, 3,... Then τ i Exp(λ) for all i, and τ i, τ j are independent for i j. In particular, E[τ i ] = λ 1 for all i. The larger the intensity λ, the more events counted by the process. 12 / 15

It is a slow day at a call center of a taxi company. There is 1 call every 10 minutes on average. Bob is the only person present and he needs to use the restroom every hour for 5 minutes. How many calls on average will he miss if he works 8 hours? What is the probability that he will not miss any call? What is the probability that he will miss more than 2 calls? Use a Poisson process to answer the questions. Let X 1, X 2,..., X 8 be the numbers of calls that arrive when Bob is in the restroom each hour. By the properties of the Poisson process, they are all independent, and all have the same distribution Pois(1/2). Hence, by the property of the Poisson distribution X = X 1 + X 2 +... + X 8 Pois(8 1/2) = Pois(4). Therefore, Bob will miss E[X] = 4 calls on average, and P(X = 0) = 40 0! e 4 = e 4 0.018, and P(X > 2) = 1 P(X 2) = = 1 40 0! e 4 41 1! e 4 42 2! e 4 = 1 (1 + 4 + 4)e 4 0.835. 13 / 15

Bob is waiting for the first call already 10 minutes. What is the probability that he will still wait more than 20 minutes? Let τ 1 be the time that Bob waits for the first call. By the property of the Poisson process τ 1 has exponential distribution with parameter 1 (here one unit is 10 minutes). We have P(τ 1 > 1 + 2 τ 1 > 1) = P(τ 1 > 2) = 1 P(τ 1 2) = e 2 0.135. 14 / 15