Department of Mathematics

Similar documents
The Longest Run of Heads

Elementary Analysis in Q p

1. INTRODUCTION. Fn 2 = F j F j+1 (1.1)

Chapter 7 Rational and Irrational Numbers

The Arm Prime Factors Decomposition

8 STOCHASTIC PROCESSES

Cryptanalysis of Pseudorandom Generators

Approximating min-max k-clustering

John Weatherwax. Analysis of Parallel Depth First Search Algorithms

1 Gambler s Ruin Problem

MATH 2710: NOTES FOR ANALYSIS

Lecture: Condorcet s Theorem

PETER J. GRABNER AND ARNOLD KNOPFMACHER

Towards understanding the Lorenz curve using the Uniform distribution. Chris J. Stephens. Newcastle City Council, Newcastle upon Tyne, UK

Almost 4000 years ago, Babylonians had discovered the following approximation to. x 2 dy 2 =1, (5.0.2)

ON THE LEAST SIGNIFICANT p ADIC DIGITS OF CERTAIN LUCAS NUMBERS

HENSEL S LEMMA KEITH CONRAD

Chapter 7: Special Distributions

ANALYTIC NUMBER THEORY AND DIRICHLET S THEOREM

1 Probability Spaces and Random Variables

Split the integral into two: [0,1] and (1, )

Lecture 3 January 16

An Overview of Witt Vectors

Statics and dynamics: some elementary concepts

Convex Optimization methods for Computing Channel Capacity

Averaging sums of powers of integers and Faulhaber polynomials

ARITHMETIC PROGRESSIONS OF POLYGONAL NUMBERS WITH COMMON DIFFERENCE A POLYGONAL NUMBER

On a Markov Game with Incomplete Information

DIRICHLET S THEOREM ON PRIMES IN ARITHMETIC PROGRESSIONS. 1. Introduction

NOTES ON RAMANUJAN'S SINGULAR MODULI. Bruce C. Berndt and Heng Huat Chan. 1. Introduction

Additive results for the generalized Drazin inverse in a Banach algebra

SOME TRACE INEQUALITIES FOR OPERATORS IN HILBERT SPACES

p-adic Measures and Bernoulli Numbers

CHAPTER-II Control Charts for Fraction Nonconforming using m-of-m Runs Rules

arxiv: v1 [math-ph] 29 Apr 2016

The non-stochastic multi-armed bandit problem

Topic 3: The Expectation of a Random Variable

Some Results on the Generalized Gaussian Distribution

Robustness of classifiers to uniform l p and Gaussian noise Supplementary material

Intrinsic Approximation on Cantor-like Sets, a Problem of Mahler

Solvability and Number of Roots of Bi-Quadratic Equations over p adic Fields

Hotelling s Two- Sample T 2

On a class of Rellich inequalities

1 Extremum Estimators

Analysis of some entrance probabilities for killed birth-death processes

A FEW EQUIVALENCES OF WALL-SUN-SUN PRIME CONJECTURE

Outline. Markov Chains and Markov Models. Outline. Markov Chains. Markov Chains Definitions Huizhen Yu

B8.1 Martingales Through Measure Theory. Concept of independence

Math 5330 Spring Notes Prime Numbers

Various Proofs for the Decrease Monotonicity of the Schatten s Power Norm, Various Families of R n Norms and Some Open Problems

Lecture 6. 2 Recurrence/transience, harmonic functions and martingales

Introduction to Probability and Statistics

New Information Measures for the Generalized Normal Distribution

Positive decomposition of transfer functions with multiple poles

Multiplicative group law on the folium of Descartes

Lecture 10: Hypercontractivity

Pell's Equation and Fundamental Units Pell's equation was first introduced to me in the number theory class at Caltech that I never comleted. It was r

A Social Welfare Optimal Sequential Allocation Procedure

Estimation of the large covariance matrix with two-step monotone missing data

UNIVERSITY OF DUBLIN TRINITY COLLEGE. Faculty of Engineering, Mathematics and Science. School of Computer Science & Statistics

MAS 4203 Number Theory. M. Yotov

p-adic Properties of Lengyel s Numbers

Collaborative Place Models Supplement 1

Understanding and Using Availability

E-companion to A risk- and ambiguity-averse extension of the max-min newsvendor order formula

A CONCRETE EXAMPLE OF PRIME BEHAVIOR IN QUADRATIC FIELDS. 1. Abstract

JEAN-MARIE DE KONINCK AND IMRE KÁTAI

Journal of Mathematical Analysis and Applications

Understanding and Using Availability

Infinitely Many Quadratic Diophantine Equations Solvable Everywhere Locally, But Not Solvable Globally

k- price auctions and Combination-auctions

REFINED STRAIN ENERGY OF THE SHELL

Department of Mathematics

Sums of independent random variables

Numerical Linear Algebra

Model checking, verification of CTL. One must verify or expel... doubts, and convert them into the certainty of YES [Thomas Carlyle]

Participation Factors. However, it does not give the influence of each state on the mode.

DIFFERENTIAL GEOMETRY. LECTURES 9-10,

1 Random Variables and Probability Distributions

ECON 4130 Supplementary Exercises 1-4

Solutions to In Class Problems Week 15, Wed.

Department of Mathematics

Moment Generating Function. STAT/MTHE 353: 5 Moment Generating Functions and Multivariate Normal Distribution

Homework Solution 4 for APPM4/5560 Markov Processes

Sharp gradient estimate and spectral rigidity for p-laplacian

0.6 Factoring 73. As always, the reader is encouraged to multiply out (3

When do Fibonacci invertible classes modulo M form a subgroup?

APPROXIMATIONS DES FORMULES

Representing Integers as the Sum of Two Squares in the Ring Z n

ε i (E j )=δj i = 0, if i j, form a basis for V, called the dual basis to (E i ). Therefore, dim V =dim V.

Generating Function Notes , Fall 2005, Prof. Peter Shor

1 Examples of Weak Induction

2016-r1 Physics 220: Worksheet 02 Name

IMPROVED BOUNDS IN THE SCALED ENFLO TYPE INEQUALITY FOR BANACH SPACES

Real Analysis 1 Fall Homework 3. a n.

Discrete Distributions Chapter 6

CERIAS Tech Report The period of the Bell numbers modulo a prime by Peter Montgomery, Sangil Nahm, Samuel Wagstaff Jr Center for Education

TRACES OF SCHUR AND KRONECKER PRODUCTS FOR BLOCK MATRICES

Why Proofs? Proof Techniques. Theorems. Other True Things. Proper Proof Technique. How To Construct A Proof. By Chuck Cusack

Outline. EECS150 - Digital Design Lecture 26 Error Correction Codes, Linear Feedback Shift Registers (LFSRs) Simple Error Detection Coding

Transcription:

Deartment of Mathematics Ma 3/03 KC Border Introduction to Probability and Statistics Winter 209 Sulement : Series fun, or some sums Comuting the mean and variance of discrete distributions often involves summing infinite series. That was the most difficult and my least favorite toic in my calculus course. Here are a few useful derivations. They aren t always clever, but they tend to follow an obvious attern, which means that even non-clever eole like me may have a hoe of re-deriving them. For a ustification of some of the oerations on infinite series of functions used, see Aostol [, Chater ]. S. Geometric series You already know this series. I am including it for the sake of comleteness. Let 0 < <. k n+ k n+ k k () (2) (3) (4) Proof : It is enough to rove (), so let x + + + n. Then simly exanding ()x yields ()x ()( + + + n ) n+, from which () follows. S.2 A weighted geometric sum S.2. Proosition For 0 < <, k k () 2. (5) It may not seem very Caltech-like to ut down cleverness. Many of you are very clever, and that is good. But relying on cleverness has a downside. My favorite comments on why one should avoid clever solutions is from the rogrammer Mark Jason Dominus in his brilliant Higher Order PERL [2,. 229]: These three tactics are resented in increasing order of cleverness. Such cleverness should be used only when necessary, since it requires a corresonding alication of cleverness on the art of the maintenance rogrammer eight weeks later, and such cleverness may not be available. KC Border v. 209.0.20::8.52

KC Border Some sums S 2 The elementary aroach: To rove (5), first fix n and let Then ()x exands to Dividing both sides by gives x + 2 2 + 3 3 + + n n ()x + 2 2 + 3 3 + + n n k k. 2 2 3 (n ) n n n+ + 2 + 3 + + n n n+ n+ n n+ by (2). k k x n+ () 2 n n+, and letting n gives as desired. k k () 2, Generating function aroach to (5): Let f() /(). For 0 < <, by (3): Differentiating term-by-term we have So multilying both sides by gives (5). f() + + 2 + () 2 f () 0 + + 2 + 3 2 + ( + 2 2 + 3 3 + ) S.3 Exected value of a geometric random variable A geometric random variable is the eoch of the first success in a sequence of indeendent reetitions of a Bernoulli trial with robability of success. (It is also a secial case of the negative binomial distribution. The mf is given by P (X k) () k, k, 2,... Rewriting () k ()k, we see that (3) imlies these robabilities sum to. To lighten the notation, let q. I claim the exectation is E X k() k. (6) v. 209.0.20::8.52 KC Border

KC Border Some sums S 3 For examle, the exected length of the St. Petersburg game (toss a coin until the first Tails) has /2, so the exected length is /(/2) 2. Proof : To rove (6), rewrite (5) by relacing with q to get kq k q ( q) 2. Multilying both sides by ( q)/q gives Now let q to get (6). k( q)q k q. S.4 An inverse exectation I claim that for a geometric X as above, E X k ()k ln ( ). (7) Proof rovided the wise TA Victor Kasatkin: Let f(q) It is analytic for q <. So for q < we may comute the derivative term-by-term: Now, f(0) 0, and thus In other words, f(q) 0 q k k. f (q) q k q q. q k k q 0 f (t) dt q 0 dt ln( q). t ( ) f(q) ln( q) ln. q Now multily both sdes by /q and relace q by to get (7) S.5 Variance of the geometric distribution If X is a geometric random variable, we can comute its variance (and higher moments). Recall that Var X E(X 2 ) (E X) 2. So let us first comute x k 2 q k. KC Border v. 209.0.20::8.52

KC Border Some sums S 4 Because of the constant of normalization, E(X 2 ) q q x. So write ( q)x k 2 q k q k 2 q k k 2 q k k 2 q k+ k 2 q k k 2 q k+ k 2 q k (k ) 2 q k ( k 2 (k ) 2) q k (2k )q k q 2 ( q) 2 q q q( + q) ( q) 2, where the last line follows from (5) and (4). The variance can now be comuted as ( q)x/q (/) 2, or Var X () 2 (8) S.6 Sums related to higher geometric moments The calculation of the variance suggested a recursive way of comuting the following series: S(n) k n q k. I don t have a lot of use for this beyond n 2, but I thought I d write it down before I forgot it. Start by writing ( q)s(n + ) k n q k k n q k+ k n q k k n q k+ k n q k (k ) n q k ( k n (k ) n) q k n ( ) n k ( ) n q k, v. 209.0.20::8.52 KC Border

KC Border Some sums S 5 where the last line is ust the Binomial Theorem. Now rearrange the terms to get or We already know that n ( q)s(n + ) n ( ) n ( ) n S(), S(n + ) q n S(0) ( ) n k ( ) n q k ( ) n ( ) n S(). q q, so with enough atience (or Mathematica) we find S(n) for any nonnegative integer n. According to Mathematica, the function S(n, q) k n q k is known as the PolyLog[-n,q] function, which can be exressed in terms of an integral over the interval [0, ]. S.7 The Taylor series for the exonential Aostol [,. 436] roves that the Taylor series for the exonential function yields the following identity. For each real number x, e x x k k!. (9) Consider the function g(x) e x. Its n th derivative is given by g (n) (x) e x, so g (n) (0) for every n, and the infinite Taylor s series exansion of g around zero is So g(x) g(0) + k! g(k) (0)(x 0) k + e x x k k!. n x k k!. S.8 Series for the logarithm When a function has reresentation asa ower series on an interval, then its indefinite integral and derivative may be found by differentiating term by term. See Theorems.8 and.9 in Aostol [,. 432]. Equation (3) tells us that the function f defined by the geometric series f() + + 2 + + n + KC Border v. 209.0.20::8.52

KC Border Some sums S 6 for <. Relacing by gives + 2 3 + + ( ) n n + +. (0) Since /( + ) d ln( + ), integrating (0) term-by-term gives for <, 2 2 + 3 3 4 4 + ( )n x n+ + ln( + ) () n + S.9 A Fun Fibonacci Sum The Fibonacci sequence is defined by the difference equation or recurrence relation with initial conditions + 2 (n > ), F 0 0, F. It can be used to define a robability distribution because To see this, observe that n 2 n+ F 4 + F 2 8 + 4 + 8 + n3 4 + 8 + n2 n3 4 + 8 + ( 4 + 8 + 2 4 + 3 4 from which it follows that Bibliograhy n n n. (2) 2n+ 2 n+ regrou 2 n+ + 2 2 n+ recursion relation n n3 F n 2 n+2 + n n 2 n+2 F 8 2 n+ F 8 + 4 2 n+3 shift indices ) + 2 n+3 regrou n n 2 n+ factor, simlify 2n+. 2n+ [] T. M. Aostol. 967. Calculus, Volume I: One-variable calculus with an introduction to linear algebra, 2d. ed. New York: John Wiley & Sons. [2] M. J. Dominus. 2005. Higher order PERL: Transforming rograms with rograms. Amsterdam: Morgan Kaufmann. v. 209.0.20::8.52 KC Border