Discrete Mathematics and Probability Theory Fall 2016 Seshia and Walrand DIS 10b

Similar documents
,m = 1,...,n; 2 ; p m (1 p) n m,m = 0,...,n; E[X] = np; n! e λ,n 0; E[X] = λ.

Lecture 02: Bounding tail distributions of a random variable

Class 13,14 June 17, 19, 2015

1. The weight of six Golden Retrievers is 66, 61, 70, 67, 92 and 66 pounds. The weight of six Labrador Retrievers is 54, 60, 72, 78, 84 and 67.

Parameter, Statistic and Random Samples

The Occupancy and Coupon Collector problems

Lecture 4 Sep 9, 2015

Random Variables and Probability Distributions

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model

Econometric Methods. Review of Estimation

å 1 13 Practice Final Examination Solutions - = CS109 Dec 5, 2018

PTAS for Bin-Packing

Lecture 3. Sampling, sampling distributions, and parameter estimation

Special Instructions / Useful Data

hp calculators HP 30S Statistics Averages and Standard Deviations Average and Standard Deviation Practice Finding Averages and Standard Deviations

Multiple Regression. More than 2 variables! Grade on Final. Multiple Regression 11/21/2012. Exam 2 Grades. Exam 2 Re-grades

Homework 1: Solutions Sid Banerjee Problem 1: (Practice with Asymptotic Notation) ORIE 4520: Stochastics at Scale Fall 2015

STATISTICAL PROPERTIES OF LEAST SQUARES ESTIMATORS. x, where. = y - ˆ " 1

Chapter 9 Jordan Block Matrices

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

THE ROYAL STATISTICAL SOCIETY HIGHER CERTIFICATE

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

CIS 800/002 The Algorithmic Foundations of Data Privacy October 13, Lecture 9. Database Update Algorithms: Multiplicative Weights

CHAPTER 4 RADICAL EXPRESSIONS

1 Solution to Problem 6.40

Simulation Output Analysis

Lecture 9: Tolerant Testing

MEASURES OF DISPERSION

Lecture 3 Probability review (cont d)

2.28 The Wall Street Journal is probably referring to the average number of cubes used per glass measured for some population that they have chosen.

Lecture Notes Types of economic variables

Chapter 4 Multiple Random Variables

STA302/1001-Fall 2008 Midterm Test October 21, 2008

CHAPTER VI Statistical Analysis of Experimental Data

Chapter 5 Properties of a Random Sample

Law of Large Numbers

arxiv:math/ v1 [math.gm] 8 Dec 2005

Continuous Distributions

Dimensionality Reduction and Learning

Qualifying Exam Statistical Theory Problem Solutions August 2005

ENGI 4421 Joint Probability Distributions Page Joint Probability Distributions [Navidi sections 2.5 and 2.6; Devore sections

THE ROYAL STATISTICAL SOCIETY 2016 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE MODULE 5

Multiple Choice Test. Chapter Adequacy of Models for Regression

Mu Sequences/Series Solutions National Convention 2014

Midterm Exam 1, section 1 (Solution) Thursday, February hour, 15 minutes

Introduction to Probability

Summary of the lecture in Biostatistics

Midterm Exam 1, section 2 (Solution) Thursday, February hour, 15 minutes

18.657: Mathematics of Machine Learning

Chapter 8: Statistical Analysis of Simulated Data

STA 105-M BASIC STATISTICS (This is a multiple choice paper.)

{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution:

Chapter 14 Logistic Regression Models

X X X E[ ] E X E X. is the ()m n where the ( i,)th. j element is the mean of the ( i,)th., then

MATH 247/Winter Notes on the adjoint and on normal operators.

Feature Selection: Part 2. 1 Greedy Algorithms (continued from the last lecture)

Section l h l Stem=Tens. 8l Leaf=Ones. 8h l 03. 9h 58

(b) By independence, the probability that the string 1011 is received correctly is

Algorithms Design & Analysis. Hash Tables

The Mathematical Appendix

CS286.2 Lecture 4: Dinur s Proof of the PCP Theorem

Ordinary Least Squares Regression. Simple Regression. Algebra and Assumptions.

TESTS BASED ON MAXIMUM LIKELIHOOD

X ε ) = 0, or equivalently, lim

Chapter 3 Sampling For Proportions and Percentages

ANSWER KEY 7 GAME THEORY, ECON 395

Hard Core Predicates: How to encrypt? Recap

The expected value of a sum of random variables,, is the sum of the expected values:

Statistics: Unlocking the Power of Data Lock 5

best estimate (mean) for X uncertainty or error in the measurement (systematic, random or statistical) best

D KL (P Q) := p i ln p i q i

Multivariate Transformation of Variables and Maximum Likelihood Estimation

For combinatorial problems we might need to generate all permutations, combinations, or subsets of a set.

n outcome is (+1,+1, 1,..., 1). Let the r.v. X denote our position (relative to our starting point 0) after n moves. Thus X = X 1 + X 2 + +X n,

ε. Therefore, the estimate

2. Independence and Bernoulli Trials

THE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA

ρ < 1 be five real numbers. The

Chapter 8. Inferences about More Than Two Population Central Values

Laboratory I.10 It All Adds Up

CS 157 Midterm Examination Fall

Arithmetic Mean and Geometric Mean

Mean is only appropriate for interval or ratio scales, not ordinal or nominal.

Logistic regression (continued)

Module 7. Lecture 7: Statistical parameter estimation

1 Onto functions and bijections Applications to Counting

Chapter 13, Part A Analysis of Variance and Experimental Design. Introduction to Analysis of Variance. Introduction to Analysis of Variance

Point Estimation: definition of estimators

STK4011 and STK9011 Autumn 2016

STA 108 Applied Linear Models: Regression Analysis Spring Solution for Homework #1

Statistics Descriptive and Inferential Statistics. Instructor: Daisuke Nagakura

Maximum Likelihood Estimation

Bayes (Naïve or not) Classifiers: Generative Approach

Likewise, properties of the optimal policy for equipment replacement & maintenance problems can be used to reduce the computation.

MA 524 Homework 6 Solutions

NP!= P. By Liu Ran. Table of Contents. The P versus NP problem is a major unsolved problem in computer

SPECIAL CONSIDERATIONS FOR VOLUMETRIC Z-TEST FOR PROPORTIONS

Lecture 2 - What are component and system reliability and how it can be improved?

Rademacher Complexity. Examples

Extend the Borel-Cantelli Lemma to Sequences of. Non-Independent Random Variables

Transcription:

CS 70 Dscrete Mathematcs ad Probablty Theory Fall 206 Sesha ad Walrad DIS 0b. Wll I Get My Package? Seaky delvery guy of some compay s out delverg packages to customers. Not oly does he had a radom package to each customer, he teds to ope a package before delverg wth probablty 2. Let X be the umber of customers who receve ther ow packages uopeed. (a) Compute the expectato E(X). (b) Compute the varace Var(X). Soluto: (a) Defe X = Hece E(X) = E( { f the -th customer gets hs/her package uopeed 0 otherwse X ) = E(X ) E(X ) = Pr[X = ] = 2 sce the -th customer wll get hs/her ow package wth probablty ad t wll be uopeed wth probablty 2 ad the delvery guy opes the packages depedetly. Hece E(X) = 2 = 2. (b) To calculate Var(X), we eed to kow E(X 2 ). By learty of expectato: E(X 2 ) = E ( (X +X 2 +...+X ) 2) = E( The we cosder two cases, ether = j or j. Hece E(X 2) = E(X X j ) = E(X 2 ) + E(X X j ) X X j ) =E(X X j ) 2 for all. To fd E(X X j ), we eed to calculate Pr[X X j = ]. Pr[X X j = ] = Pr[X = ]Pr[X j = X = ] = 2 2( ) sce f customer has receved hs/her ow package, customer j has choces left. Hece E(X 2 ) = 2 + ( ) 2 2( ) = 3 4. Var(X) = E(X 2 ) E(X) 2 = 3 4 4 = 2. 2. Varace Ths problem wll gve you practce usg the stadard method to compute the varace of a sum of radom varables that are ot parwse depedet (so you caot use learty of varace). CS 70, Fall 206, DIS 0b

(a) A buldg has floors umbered,2,...,, plus a groud floor G. At the groud floor, m people get o the elevator together, ad each gets off at a uformly radom oe of the floors (depedetly of everybody else). What s the varace of the umber of floors the elevator does ot stop at? (I fact, the varace of the umber of floors the elevator does stop at must be the same, but the former s a lttle easer to compute.) (b) A group of three freds has books they would all lke to read. Each fred (depedetly of the other two) pcks a radom permutato of the books ad reads them that order, oe book per week (for cosecutve weeks). Let X be the umber of weeks whch all three freds are readg the same book. Compute Var(X). Soluto: (a) Let X be the umber of floors the elevator does ot stop at. As the prevous homework, we ca represet X as the sum of the dcator varables X,...,X, where X = f o oe gets off o floor. Thus, we have ( ) m E(X ) = Pr[X = ] =, ad from learty of expectato, E(X) = ( ) m E(X ) =. To fd the varace, we caot smply sum the varace of our dcator varables. However, we ca stll compute Var(X) = E(X 2 ) E(X) 2 drectly usg learty of expectato, but ow how ca we fd E(X 2 )? Recall that E(X 2 ) = E((X +... + X ) 2 ) = E(X X j ) = E(X X j ) = E(X 2 ) + E(X X j ). The frst term s smple to calculate: E(X 2) = 2 Pr[X = ] = ( ) m, meag that ( ) m E(X 2 ) =. X X j = whe both X ad X j are, whch meas o oe gets off the elevator o floor ad floor j. Ths happes wth probablty ( ) 2 m Pr[X = X j = ] = Pr[X = X j = ] =. CS 70, Fall 206, DIS 0b 2

Thus, we ca ow compute ( ) 2 m E(X X j ) = ( ). Fally, we plug to see that ( ) m ( ) 2 m ( ( ) m ) 2 Var(X) = E(X 2 ) E(X) 2 = + ( ). (b) Let X,...,X be dcator varables such that X = f all three freds are readg the same book o week. Thus, we have ( ) 2 E(X ) = Pr[X = ] =, ad from learty of expectato, As before, we kow that E(X) = E(X 2 ) = E(X ) = ( ) 2 =. E(X 2 ) + E(X X j ). Furthermore, because X s a dcator varable, E(X 2) = 2 Pr[X = ] = ( ) 2, ad ( ) 2 E(X 2 ) = =. Aga, because X ad X j are dcator varables, we are terested Pr[X = X j = ] = Pr[X = X j = ] = (( )) 2, the probablty that all three freds pck the same book o week ad week j. Thus, ( ) E(X X j ) = ( ) (( )) 2 = ( ). Fally, we compute Var(X) = E(X 2 ) E(X) 2 = ( ) + 2 ( ). 3. Markov s Iequalty ad Chebyshev s Iequalty A radom varable X has varace Var(X) = 9 ad expectato E(X) = 2. Furthermore, the value of X s ever greater tha 0. Gve ths formato, provde ether a proof or a couterexample for the followg statemets. CS 70, Fall 206, DIS 0b 3

(a) E(X 2 ) = 3. (b) Pr[X = 2] > 0. (c) Pr[X 2] = Pr[X 2]. (d) Pr[X ] 8/9. (e) Pr[X 6] 9/6. (f) Pr[X 6] 9/32. Soluto: (a) TRUE. Sce 9 = Var(X) = E(X 2 ) E(X) 2 = E(X 2 ) 2 2, we have E(X 2 ) = 9+4 = 3. (b) FALSE. Costruct a radom varable X that satsfes the codtos the questo but does ot take o the value 2. A smple example would be a radom varable that takes o 2 values, where Pr[X = a] = 2,Pr[X = b] = 2, ad a b. The expectato must be 2, so we have 2 a + 2 b = 2. The varace s 9, so E(X 2 ) = 3 (from part (a)) ad 2 a2 + 2 b2 = 3. Solvg for a ad b, we get Pr[X = ] = 2,Pr[X = 5] = 2 as a couterexample. (c) FALSE. Costruct a radom varable X that satsfes the codtos the questo but does ot have a equal chace of beg less tha or greater tha 2. A smple example would be a radom varable that takes o 2 values, where Pr[X = a] = p,pr[x = b] = p. Here, we use the same approach as part (b) except wth a geerc p, sce we wat p. The expectato must be 2, so we have pa + ( p)b = 2. The varace s 9, so E(X 2 ) = 3 ad pa 2 + ( p)b 2 = 3. Solvg for a ad b, we fd the relato b = 2± 3 x, where x = p p. The, we ca fd a example by pluggg values for x so that a,b 0 ad p 2. Oe such couterexample s Pr[X = 7] = 0 9,Pr[X = 3] = 0. (d) TRUE. Let Y = 0 X. Sce X s ever exceeds 0, Y s a o-egatve radom varable. By Markov s equalty, Pr[0 X a] = Pr[Y a] E(Y ) a Settg a = 9, we get Pr[X ] = Pr[0 X 9] 8 9. = E(0 X) a = 8 a. (e) TRUE. Chebyshev s equalty says Pr[ X E[X] a] Var(X). If we set a = 4, we a 2 have Pr[ X 2 4] 9 6. Now we smply observe that Pr[X 6] Pr[ X 2 4], because the evet X 6 s a subset of the evet X 2 4. (f) FALSE. We use the same approach as part (c), except we fd a couterexample that fts the equalty Pr[X 6] 9/32. Oe example s Pr[X = 0] = 9 3,Pr[X = 3 2 ] = 4 3. CS 70, Fall 206, DIS 0b 4

4. Easy A s A fred tells you about a course called Lazess Moder Socety that requres almost o work. You hope to take ths course ext semester to gve yourself a well-deserved break after masterg CS70. At the frst lecture, the professor aouces that grades wll deped oly a mdterm ad a fal. The mdterm wll cosst of three questos, each worth 0 pots, ad the fal wll cosst of four questos, also each worth 0 pots. He wll gve a A to ay studet who gets at least 60 of the possble 70 pots. However, speakg wth the professor offce hours you hear some very dsturbg ews. He tells you that, the sprt of the class, the GSIs are very lazy, ad to save tme the gradg wll be doe as follows. For each studet s mdterm, the GSIs wll choose a real umber radomly from a ormal dstrbuto wth mea µ = 5 ad varace σ 2 =. They ll mark each of the three questos wth that score. To grade the fal, they ll aga choose a radom umber from the same dstrbuto, depedet of the frst umber, ad wll mark all four questos wth that score. If you take the class, what wll the mea ad varace of your total class score be? Use Chebyshev s equalty to coclude that you have less tha a 5% chace of gettg a A. Soluto: Let X be the total umber of pots you receve the class. The X = X m + X f where X m are the pots you receve o mdterm ad X f are the pots you receve o the fal. Your mdterm score s geerated as X m = 3Y m, where the r.v. Y m represets the real umber that the professor chose whe gradg your mdterm. Smlarly, X f = 4Y f. The problem statemet tells us that Y m Normal(5,) ad Y f Normal(5,), so E[Y m ] = E[Y f ] = 5 ad Var(Y m ) = Var(Y f ) =. Thus, E[X] = E[X m ] + E[X f ] = 3E[Y m ] + 4E[Y f ] = 35 ad Var(X) = Var(X m ) + Var(X f ) = 9Var(Y m ) + 6Var(Y f ) = 25. Usg Chebyshev s Iequalty, we get Pr[X 60] Pr[ X 35 25] Var(X) 25 2 = 25. Ufortuately, you have at most a 4% chace of gettg a A. So, the aswer s: your mea score wll be 35, the varace wll be 25, ad yes, you ca coclude that you have less tha a 5% chace of gettg a A. Note that although we calculated a boud for Pr[ X 35 25], whch s the probablty that you wll get 60 or above or 0 or below, we caot dvde by 2 to refe our boud uless the dstrbuto s symmetrc about ts mea. I ths case, the dstrbuto s ot symmetrc. CS 70, Fall 206, DIS 0b 5