STAT 516 Answers Homework 6 April 2, 2008 Solutions by Mark Daniel Ward PROBLEMS

Similar documents
Problems from 9th edition of Probability and Statistical Inference by Hogg, Tanis and Zimmerman:

Randomized Algorithms I, Spring 2018, Department of Computer Science, University of Helsinki Homework 1: Solutions (Discussed January 25, 2018)

Final Review for MATH 3510

STAT Homework 1 - Solutions

Exercise 4.3 Use the Continuity Theorem to prove the Cramér-Wold Theorem, Theorem. (1) φ a X(1).

EE 4TM4: Digital Communications II Probability Theory

4. Partial Sums and the Central Limit Theorem

7.1 Convergence of sequences of random variables

Limit Theorems. Convergence in Probability. Let X be the number of heads observed in n tosses. Then, E[X] = np and Var[X] = np(1-p).


Discrete Mathematics and Probability Theory Spring 2016 Rao and Walrand Note 19

IE 230 Seat # Name < KEY > Please read these directions. Closed book and notes. 60 minutes.

Chapter 6 Principles of Data Reduction

PRACTICE PROBLEMS FOR THE FINAL

This exam contains 19 pages (including this cover page) and 10 questions. A Formulae sheet is provided with the exam.

1 Generating functions for balls in boxes

Discrete Mathematics for CS Spring 2005 Clancy/Wagner Notes 21. Some Important Distributions

IIT JAM Mathematical Statistics (MS) 2006 SECTION A

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,

0, otherwise. EX = E(X 1 + X n ) = EX j = np and. Var(X j ) = np(1 p). Var(X) = Var(X X n ) =

f X (12) = Pr(X = 12) = Pr({(6, 6)}) = 1/36

7.1 Convergence of sequences of random variables

Lecture 7: Properties of Random Samples

Mathematical Statistics - MS

CS/ECE 715 Spring 2004 Homework 5 (Due date: March 16)

Approximations and more PMFs and PDFs

Lecture 18: Sampling distributions

CS 330 Discussion - Probability

1 Approximating Integrals using Taylor Polynomials

NOTES ON DISTRIBUTIONS

HOMEWORK I: PREREQUISITES FROM MATH 727

6.3 Testing Series With Positive Terms

Random Variables, Sampling and Estimation

Lecture 2: Monte Carlo Simulation

STAT Homework 2 - Solutions

Distribution of Random Samples & Limit theorems

Unbiased Estimation. February 7-12, 2008

CS284A: Representations and Algorithms in Molecular Biology

Discrete probability distributions

Continuous Random Variables: Conditioning, Expectation and Independence

Quick Review of Probability

1 Introduction to reducing variance in Monte Carlo simulations

Quick Review of Probability

Solutions to Homework 6

4.3 Growth Rates of Solutions to Recurrences

Chapter 2 The Monte Carlo Method

Probability and statistics: basic terms

Simulation. Two Rule For Inverting A Distribution Function

f X (12) = Pr(X = 12) = Pr({(6, 6)}) = 1/36

The Discrete-Time Fourier Transform (DTFT)

Statistical Signal Processing

Discrete Mathematics and Probability Theory Spring 2012 Alistair Sinclair Note 15

EECS564 Estimation, Filtering, and Detection Hwk 2 Solns. Winter p θ (z) = (2θz + 1 θ), 0 z 1

Homework 5 Solutions

PRACTICE PROBLEMS FOR THE FINAL

It is always the case that unions, intersections, complements, and set differences are preserved by the inverse image of a function.

McGill University Math 354: Honors Analysis 3 Fall 2012 Solutions to selected problems. x y

Probability 2 - Notes 10. Lemma. If X is a random variable and g(x) 0 for all x in the support of f X, then P(g(X) 1) E[g(X)].

Probability theory and mathematical statistics:

Infinite Sequences and Series

Discrete Mathematics for CS Spring 2007 Luca Trevisan Lecture 22

Mathematics 170B Selected HW Solutions.

Discrete Mathematics and Probability Theory Summer 2014 James Cook Note 15

IE 230 Probability & Statistics in Engineering I. Closed book and notes. No calculators. 120 minutes.

Hypothesis Testing. Evaluation of Performance of Learned h. Issues. Trade-off Between Bias and Variance

AMS570 Lecture Notes #2

Expectation and Variance of a random variable

Massachusetts Institute of Technology

x 2 x x x x x + x x +2 x

The Poisson Process *

TMA4245 Statistics. Corrected 30 May and 4 June Norwegian University of Science and Technology Department of Mathematical Sciences.

Lecture 1 Probability and Statistics

f(x) dx as we do. 2x dx x also diverges. Solution: We compute 2x dx lim

FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING. Lectures

Topic 9: Sampling Distributions of Estimators

Stat 400: Georgios Fellouris Homework 5 Due: Friday 24 th, 2017

EXAMINATIONS OF THE ROYAL STATISTICAL SOCIETY

6. Sufficient, Complete, and Ancillary Statistics

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 2 9/9/2013. Large Deviations for i.i.d. Random Variables

CS321. Numerical Analysis and Computing

FIR Filter Design: Part I

Math 10A final exam, December 16, 2016

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS

An Introduction to Randomized Algorithms

CS537. Numerical Analysis and Computing

PUTNAM TRAINING PROBABILITY

Economics 241B Relation to Method of Moments and Maximum Likelihood OLSE as a Maximum Likelihood Estimator

AMS 216 Stochastic Differential Equations Lecture 02 Copyright by Hongyun Wang, UCSC ( ( )) 2 = E X 2 ( ( )) 2

Exponential Families and Bayesian Inference

CS / MCS 401 Homework 3 grader solutions

2 Definition of Variance and the obvious guess

Discrete Mathematics for CS Spring 2008 David Wagner Note 22

UNIT 2 DIFFERENT APPROACHES TO PROBABILITY THEORY

First Year Quantitative Comp Exam Spring, Part I - 203A. f X (x) = 0 otherwise

Convergence of random variables. (telegram style notes) P.J.C. Spreij

Kurskod: TAMS11 Provkod: TENB 21 March 2015, 14:00-18:00. English Version (no Swedish Version)

Math 2784 (or 2794W) University of Connecticut

STAT 350 Handout 19 Sampling Distribution, Central Limit Theorem (6.6)

Lesson 10: Limits and Continuity

Confidence Intervals for the Population Proportion p

Transcription:

STAT 56 Aswers Homework 6 April 2, 28 Solutios by Mark Daiel Ward PROBLEMS Chapter 6 Problems 2a. The mass p(, correspods to either o the irst two balls beig white, so p(, 8 7 4/39. The mass p(, correspods to the irst ball beig red ad the secod ball 3 2 beig white, so p(, 8 5 /39. The mass p(, correspods to the irst ball beig 3 2 white ad the secod ball beig red, so p(, 5 8 /39. The mass p(, correspods 3 2 to both o the irst two balls beig white, so p(, 5 4 5/39. 3 2 2b. Similarly, we have p(,, 8 7 6 28, ad p(,, p(,, p(,, 3 2 43 8 7 5 8 5 4, ad p(,, p(,, p(,,, ad ially p(,, 7 3 2 429 5 4 3 5. 3 2 43 4 3 2 729 3a. The mass p(, correspods to the white balls umbered ad 2 ot appearig withi the three choices, so p(, ( 3 5/26. The mass p(, correspods to the white ( 3 3 balls umbered ot beig chose, ad the white ball umbered 2 gettig chose, withi the three choices, so p(, ( ( 2 5/26. The mass p(, correspods to the white balls ( 3 3 umbered 2 ot beig chose, ad the white ball umbered gettig chose, withi the three choices, so p(, ( ( 2 5/26. The mass p(, correspods to both o the white ( 3 3 balls umbered ad 2 appearig withi the three choices, so p(, (2 2( /26. ( 3 3 3b. Similarly, we have p(,, ( 3 ( 3 3 6, ad p(,, p(,, p(,, 43 ( ( 2 ( 3 3 45, ad p(,, p(,, p(,, ( ( ( 286 ( 3 3 5, ad ially p(,, 43 ( ( ( ( 3 3 286. 5. The oly modiicatio rom problem 3a above is that the balls are replaced ater each draw. Thus p(, ( 3 ( 3 33/297; also, p(, 3 2 ( ( 3 3 + 3 ( 2 ( 3 3 + 3 3 397/297; similarly, p(, 397/297; ad ially, p(, 6 ( ( ( ( 3 3 3 +3 ( 2 3 3 + 3 ( 3 2 ( 3 72/297. 6. We irst ote that N, N 2 4 ad N + N 2 5. There are ( 5 2 ways that the two deective items ca be chose, each equally likely, ad each way correspods to exactly oe pair, 2 satisyig the bouds metioed. Thus p(, 2 / i, 2 4 ad + 2 5. Otherwise p(, 2.

2 a. We irst compute Aother method is to compute P (X < Y P (X < Y y x e (x+y dx dy /2 e (x+y dy dx /2 Fially, a third method is to just otice that X ad Y are idepedet ad have the same distributio, so hal the time we have X < Y ad the other hal o the time we have Y < X. Thus P (X < Y /2. b. For a <, we have P (X < a. For a >, we compute P (X < a a (x, y dx dy a e (x+y dx dy e a Aother method is to simply ote that the joit desity o X ad Y shows us that, i this case, X ad Y must be idepedet expoetial radom variables, each with λ. So P (X < a e a or a >, ad P (X < a otherwise, sice this is the cumulative distributio uctio o a expoetial radom variable.. There are ( 5 2,,2 3 ways that the customers ca be split ito a pair who will buy ordiary sets, a sigle buyer who will buy plasma, ad a pair who will buy othig. The probability o each such choice is (.45 2 (.5 (.4 2.486. So the desired probability is (3(.486.458. 4. We write X or the locatio o the ambulece, ad Y or the locatio o the accidet, both i the iterval [, L]. The distace i betwee is D X Y. We kow that P (D < a or a < ad P (D < a or a L, sice the miimum ad maximum distaces or D are ad L, respectively. So D must be betwee ad L. Perhaps the easiest method or computig P (D < a with a L is to draw a picture o the sample space ad the divide the desired area over the etire area o the sample space; this method works sice the joit distributio o X, Y is uiorm. So the desired probability is 2 (L a2 L 2 2 (L a2 a(2l a/l 2. L 2 Aother possibility is to itegrate, ad we eed to break the desired itegral ito three regios: P (D < a a x+a dy dx + L2 L a x+a a x a L L dy dx + L2 L a x a dy dx a(2l a/l2 L2 5a. We ote that c dx dy, so /c dx dy area o regio R. R R 5b. We see that, or < a, b <, we have F X,Y (a, b P (X a, Y b a b a ( b ( /4 dy dx. Thus F 2 2 X,Y (x, y x ( y (. So we have successully actored the cumulative distributio uctio ito two parts: oe part is a uctio o 2 2 x ad the other part is a uctio o y, so X ad Y are idepedet, ad their cumulative distributio uctios have the required orms or uiorm radom variables o the iterval (,, i.e., F X (a F Y (a a ( or < a < ad F 2 X (a F Y (a otherwise. 5c. We itegrate X 2 +Y 2 (/4 dx dy area o the circle X2 + Y 2 area o the etire square π2 π/4. 4 6a. We see that A happes i ad oly i at least oe o the A i happe. So A i A i.

6b. Yes, the A i are mutually exclusive. We caot have more tha oe A i occur at the same time. 6c. Sice the A i are mutually exclusive, the P (A P ( i A i i P (A i. We see that P (A i /2, sice we require each o the other poits (besides the ith poit itsel, o course to be i the semicircle clockwise o the ith poit. So P (A i /2. 2 9a. The margial desity o X is X (x or x ad also or x. For < x <, the margial desity o X is x X (x x dy 9b. The margial desity o Y is Y (y or y ad also or y. For < y <, the margial desity o Y is 9c. The expected value o X is E[X] 9c. The expected value o Y is E[Y ] Y (y y x X (x dx y Y (y dy dx l(/y x (x( dx /2 y l(/y dy /4 To see this, use itegratio by parts, with u l(/y ad dv y dy. 2a. Yes, X ad Y are idepedet, because we ca actor the joit desity as ollows: (x, y X (x Y (y, where { { xe x x > e y y > X (x Y (y else else 2b. No; i this case, X ad Y are ot idepedet. To see this, we ote that the desity is ozero whe < x < y <. So the domai does ot allow us to actor the joit desity ito two separate regios. For istace, P ( < X < > sice X ca be i the rage 4 betwee /4 ad. O the other had, P ( < X < 4 Y 8, sice X caot be i the rage betwee /4 ad whe Y /8; istead, X must always be smaller tha Y. 23a. Yes, X ad Y are idepedet, because we ca actor the joit desity as ollows: (x, y X (x Y (y, where { { 6x( x < x < 2y < y < X (x Y (y else else 23b. We compute E[X] x X(x dx x6x( x dx /2. 23c. We compute E[Y ] y Y (y dy y2y dy 2/3. 23d. We compute E[X 2 ] x2 X (x dx x2 6x( x dx 3/. Thus Var(X 3 ( 2 2 /2. 3

4 23e. We compute E[Y 2 ] y2 Y (y dy y2 2y dy /2. Thus Var(Y 2 ( 2 3 2 /8. 25. Sice N is Biomial with 6 ad p / 6, the is large ad p is small, so N is well approximated by a Poisso radom variable with λ p. So P (N i e λ λ i e. i! i! 26a. Sice A, B, C are idepedet, we multiple their margial desities to get the joit desity. Each o these variables has desity o the iterval (, ad desity otherwise. So the joit desity is (a, b, c or < a, b, c < ad (a, b, c. So the joit distributio is F (a, b, c F A (af B (bf C (c, where F A (a, F B (b, ad F C (c are each the cumulative distributio uctios o a uiorm (, radom variable, i.e., each o these uctios has the orm F (x i x, or F (x x i < x <, or F (x i x. 26b. The roots o the equatio Ax 2 +Bx+C are give by x b± b 2 4ac; these roots are 2a real i ad oly i b 2 4ac, which happes with probability (a, b, c da db dc, b 2 4ac which is exactly the volume o the regio {(a, b, c b 2 4ac } divided by the volume o the etire regio {(a, b, c < a, b, c < }. The secod regio has volume. The irst regio has volume mi{,b 2 /(4c} dadbdc.2544. So the desired probability is approximately.2544. To see how to do the itegral above, we compute mi{,b 2 /(4c} } da db dc mi {, b2 db dc 4c which simpliies to ( /4 4c b 2 4c db + b 2 db dc + 4c /4 4c db dc 5 36 + l(2.2544 6 28. The cumulative distributio uctio o Z X /X 2 is P (Z a or a, sice Z is ever egative i this problem. For < a, we compute ( ax2 X P (Z a P a P (X ax 2 λ λ 2 e (λ x +λ 2 x 2 dx dx 2 λ a X 2 λ a + λ 2 A alterative method o computig is to write ( ( X P (Z a P a P X 2 a X X 2 x /a λ λ 2 e (λ x +λ 2 x 2 dx 2 dx λ a λ a + λ 2 32a. We assume that the weekly sales i separate weeks is idepedet. Thus, the umber o the mea sales i two weeks is (by idepedece simply (2(22 44. The variace o sales i oe week is 23 2, so that variace o sales i two weeks is (by idepedece simply (2(23 2 5,8. So the sales i two weeks, deoted by X, has ormal distributio with mea 44 ad variace 5,8. So P (X > 5 P ( X 44 5,8 > P (Z >.84 P (Z.84 5 44 5,8

5 Φ(.84.967.329 32b. The weekly sales Y has ormal distributio with mea 22 ad variace 23 2 52,9. So, i a give week, the probability p that the weekly sales Y exceeds 2 is p P (Y > 2 ( Y 22 P > 52,9 P (Z >.87 P (Z <.87 Φ(.87.878 2 22 52,9 The ( probability that weekly sales exceeds 2 i at least 2 out o 3 weeks is (approximately 3 2 p 2 ( p + ( 3 3 p 3.934. 33a. Write X or Jill s bowlig scores, so X is ormal with mea 7 ad variace 2 2 4. Write Y or Jack s bowlig scores, so Y is ormal with mea 6 ad variace 5 2 225. So X is ormal with mea 7 ad variace 2 2 4. Thus, Y X is omal with mea 6 7 ad variace 225 + 4 625. So the desired probability is approximately ( Y X ( P (Y X > P > ( 625 625 ( P Z > 2 5 ( P Z 2 5 Φ(.4.6554.3446 Sice the bowlig scores are actually discrete iteger values, we get a eve better approximatio by usig cotiuity correctio P (Y X > P (Y X.5 ( Y X ( P > 625 P (Z >.42 P (Z.42 Φ(.42.6628.3372.5 ( 625

6 33b. The total o their scores, X + Y, is omal with mea 6 + 7 33 ad variace 225 + 4 625. So the desired probability is approximately ( X + Y 33 35 33 P (X + Y > 35 P > 625 625 ( P Z > 4 5 P (Z.8 Φ(.8.788.29 Sice the bowlig scores are actually discrete iteger values, we get a eve better approximatio by usig cotiuity correctio ( X + Y 33 35.5 33 P (X + Y 35.5 P > 625 625 P (Z >.82 P (Z.82 Φ(.82.7939.26 37a. We recall rom problem 3 that Y ad Y 2 have joit mass p(, 5/26, p(, 5/26, p(, 5/26, ad p(, /26. 5/26 So the coditio mass o Y, give that Y 2, is p Y Y 2 ( 5 ad 5/26+/26 6 p Y Y 2 ( /26 5/26+/26 6. 37b. The coditio mass o Y, give that Y 2, is p Y Y 2 ( p Y Y 2 ( 5/26 5/26+5/26 4. 39a. For y x 5, the joit mass is p(x, y p(xp(y x 5 39b. The coditio mass o X, give Y i, is p X Y (x i p(x, i p Y (i p(x, i 5 x p(x, i 5x 5 x 5x. x 5x 5x 37/3 6 37x 5/26 5/26+5/26 3 4 ad 4. First, ote that there is oly oe way to obtai X ad Y as the same value, but or X > Y, there are two ways to obtai X ad Y as the same value. So, or y < i, the coditioal mass o Y, give X i, is p Y X (y i p(i, y p X (i p(i, y i y p(i, y (2(/6(/6 (i (2(/6(/6 + (/6(/6 2 2i For y i, the coditioal mass o Y, give X i, is p Y X (y i p(i, y p X (i p(i, y i y p(i, y (/6(/6 (i (2(/6(/6 + (/6(/6 2i

For y > i, the coditioal mass o Y, give X i, is p Y X (y i. Also ote that X ad Y are depedet. For istace, P (Y > 3, because Y ca take the value 3. O the other had P (Y > 3 X 2. So X ad Y are depedet. Oce X is give, or istace, the Y ca be o larger tha X. 4a. The coditioal mass uctio o X give Y is p X Y (, p(, p Y ( 8 8 + 8 /2 ad p X Y (2, The coditioal mass uctio o X give Y 2 is p X Y (, 2 p(, 2 p Y (2 4 4 + 2 /3 ad p X Y (2, 2 p(2, p Y ( 8 8 + 8 p(2, 2 p Y (2 2 4 + 2 /2 2/3 4b. Sice the coditioal mass o X chages depedig o the value o Y, the the value o Y aects the various probabilities or X, so X ad Y are ot idepedet. 4c. We compute 7 ad ad P (XY 3 p(, + p(2, + p(, 2 8 + 8 + 4 2 P (X + Y > 2 p(2, + p(, 2 + p(2, 2 8 + 4 + 2 7 8 P (X/Y > p(2, /8 54a. We see that u g (x, y xy ad v g 2 (x, y x/y. Thus x h (u, v uv ad y h 2 (u, v u. The Jacobia is v J(x, y y x x y x y 2x y so J(x, y y. Thereore the joit desity o U, V is 2x U,V (u, v X,Y (x, y J(x, y y x 2 y 2 2x 2x 3 y 2( uv 3 u v y x y 2 2u 2 v 57. We see that y g (x, x 2 x + x 2 ad y 2 g 2 (x, x 2 e x. Thus x h (y, y 2 l(y 2 ad x 2 y l(y 2. The Jacobia is J(x, y e x ex so J(x, y e x. Thereore the joit desity o Y, Y 2 is Y,Y 2 (y, y 2 X,X 2 (x, x 2 J(x, x 2 λ λ 2 e λ x λ 2 x 2 e x λ λ 2 e ((λ +x +λ 2 x 2 λ λ 2 e ((λ + l(y 2 +λ 2 (y l(y 2 λ λ 2 y λ +λ 2 2 e y λ 2

8 THEORETICAL EXERCISES 2. Geeralizig Propositio 2. (there is othig special about two variables, we ote that radom discrete variables X,..., X are idepedet i ad oly i their joit mass (x,..., x ca be actored as (x (x ; i this case, oce each i is ormalized so that x i i (x i or each i, the the i s are the margial mass uctios o the X i s. Write X or the total umber o evets i the give time period, ad write X i as the umber o evets o type i. The we ca actor the joit mass (x,..., x o X,..., X by writig (x,..., x P (X x + + x P (X x,..., X x X + + X e λ λ x + +x ( x + + x p x p x (x + + x! x,..., x e λ λ x + +x (x + + x! p x p x (x + + x!x! x! e λ λ x λ x p x p x x! x! e λp (λp x x! e λp (λp x x! So e λp i(λp i x i x i! is the mass o X i, ad also X,..., X are idepedet. 5a. For a >, the cumulative distributio uctio o Z is F Z (a P (Z a P (X/Y a P (X ay ay X (x Y (y dx dy Y (y ay X (x dx dy Y (yf X (ay dy, or equivaletly, or z >, we have F Z (z Dieretiatig throughout with respect to z yields Z (z Y (yf X (zy dy Y (y X (zyy dy O course, or z, we have Z (z. Whe X, Y are idepedet expoetial radom variables with parameters λ, λ 2, this yields Z (z λ 2 e λ 2y λ e λ zy y dy λ λ 2 (λ 2 + zλ 2 or z > ; as beore, Z (z or z. 5b. For a >, the cumulative distributio uctio o Z is F Z (a P (Z a P (XY a P (X a/y a/y X (x Y (y dx dy Y (y a/y X (x dx dy Y (yf X (a/y dy, or equivaletly, or z >, we have F Z (z Y (yf X (z/y dy

Dieretiatig throughout with respect to z yields Z (z Y (y X (z/y y dy O course, or z, we have Z (z. Whe X, Y are idepedet expoetial radom variables with parameters λ, λ 2, this yields Z (z λ 2 e λ 2y λ e λ z/y y dy or z >, but I do ot see a easy way to simpliy this expressio; as beore, Z (z or z. 6. Method. My solutio does NOT use iductio. I the X i are idepedet ad idetically distributed geometrics, each with probability p o success, the P (X + + X i (( p x p (( p x p x + +x i p ( p x + +x i p ( p x + +x i ( p x + +x ( p i p ( p ( pi ( p i ( pi ( p ( i p ( p i x + +x i usig Propositio 6. o Chapter ad thus X + + X has a egative biomial distributio with parameters p ad. Method 2. Here is a solutio with o derivatio eeded. Do cosecutive experimets. I each experimet, lip a coi with probability p o ladig heads. The X + + X is the total umber o lips required. O course, the total umber o lips is a egative biomial radom variable, because the experimet eds whe we see the th success, or i other words, whe we see the appearace o the th head. 9. For coveiece, write Y mi(x,..., X. For a, we kow that F Y (a, sice Y is ever egative. For a >, we have F Y (a P (Y a P (Y > a P (mi(x,..., X > a P (X > a, X 2 > a,..., X > a P (X > ap (X 2 > a P (X > a (e λa (e λa (e λa e λa. Thus Y is expoetially distributio with parameter λ.. The lashlight begis with 2 batteries istalled ad 2 replacemet batteries available. Whe oe battery dies, the battery is immediately replaced, ad because o the memoryless property o expoetial radom variables, both batteries istalled (the old ad ew are just as good as ew, ad the waitig begis agai. A dead battery is replaced a total o 2 times. Upo the ( st battery dyig, we do ot have eough batteries let to ru the lashlight aymore. So the legth o time that the lashlight ca operate is X + + X, 9

where the X i s are idepedet, idetically distributio expoetial radom variables, each with parameter λ. [So the legth o time that the lashlight ca operate is a gamma radom variable with parameters (, λ; see, or istace, Example 3b o page 282. I you did ot get this last setece, do t worry, because you are ot required to uderstad gamma radom variables i my course.] 4a. We thik o X as the umber o lips o a coi util the irst appearace o heads, ad Y as the umber o additioal lips o a coi util the secod appearace o heads; here, heads appears o each toss with probability p. Give that X + Y, we kow that lips were required to reach the secod head. So ay o the irst lips could be the irst head, ad all such possibilities are equally likely, so P (X i X + Y or i ad P (X i X + Y otherwise. 4b. To veriy this, we ote that i i or i, we must have P (X i X +Y, because X, Y are positive radom variables. For i, we compute 5. Method. We compute P (X i ad X + Y P (X i X + Y P (X + Y P (X i, Y i P (X + Y P (X ip (Y i j P (X jp (Y j ( p i p( p i p j ( pj p( p j p ( p 2 ( ( p 2 P (X i ad X + Y m P (X i X + Y m P (X + Y m P (X i, Y m i P (X + Y m P (X ip (Y m i m j P (X jp (Y m j ( i p i ( p i( m i p m i ( p m+i m ( j j pj ( p j( m j p m j ( p m+j ( i( m i ( j( m j ( i( m i ( 2 m m j

Method 2. As i Ross s hit, lip 2 cois, let X deote the umber o heads i the irst sequece o lips ad let Y deote the umber o heads i the secod sequece o lips. I m lips are heads altogether, there are ( ( 2 m equally likely possibilities, exactly ( i m i o which have i heads i the irst sequece o lips ad the other m i heads i the secod sequece o lips. So P (X i X + Y m ( i( m i, ad thus, give X + Y m, we ( 2 m see that the coditioal distributio o X is hypergeometric. 8a. Give the coditio U > a, the U is uiormly distributed o the iterval (a,. To see this, just cosider ay b with a < b <. The P (U < b U > a P (a<u<b b a. P (U>a a Dieretiatig with respect to b yields the coditioal desity o U, amely,, which is a costat (sice a is ixed i this problem. So the coditioal distributio o U is uiorm o the iterval (a,, as we stated at the start. 8b. Give the coditio U < a, the U is uiormly distributed o the iterval (, a. To see this, just cosider ay b with < b < a. The P (U < b U < a P (U<b b. P (U<a a Dieretiatig with respect to b yields the coditioal desity o U, amely,, which is a costat (sice a is ixed i this problem. So the coditioal distributio o U is uiorm o the iterval (, a, as we stated at the start.