SOLUTIONS TO SELF-QUIZZES AND SELF-TESTS CHAPTER 10 FUNDAMENTALS OF PROBABILITY WITH STOCHASTIC PROCESSES FOURTH EDITION SAEED GHAHRAMANI Western New England University Springfield, Massachusetts, USA A CHAPMAN & HALL BOOK Chapter 10 Solutions to Self-Quiz Problems 2 SOLUTIONS TO SELF-QUIZ PROBLEMS Section 10.1 1. Let X be the number of disk storage wallets that Shiante should search to find the DVD. For i = 1, 2, . . ., 11, let Xi = 1 if the DVD is stored in the ith wallet that Shiante searches, and Xi = 0, otherwise. Then X = 1 · X1 + 2 · X2 + · · · + 11 · X11 , and, therefore, E(X) = 1 · E(X1) + 2 · E(X2) + · · · + 11 · E(X11) =1· 1 1 1 1 11 × 12 1 +2· + · · · + 11 · = (1 + 2 + · · · + 11) = · = 6. 11 11 11 11 11 2 2. By the Cauchy-Schwarz inequality, E(XY ) ≤ E(X 2)E(Y 2 ) . Note that XY = X and Y 2 = Y . Since X is nonnegative, and avoiding trivial cases, E(X) > 0, we have E(X) = E(X) = E(XY ) ≤ E(X 2)E(Y 2 ). However, E(Y 2 ) = E(Y ) = P (X > 0). Thus E(X) ≤ This implies that 2 E(X 2)P (X > 0) ≥ E(X) . Hence or E(X 2)P (X > 0). 2 E(X) , P (X > 0) ≥ E(X 2) 2 E(X) , 1 − P (X > 0) ≤ 1 − E(X 2) which establishes the relation 2 E(X 2) − E(X) Var(X) = . P (X = 0) ≤ 2 E(X ) E(X 2) Chapter 10 Solutions to Self-Quiz Problems 3 Section 10.2 1. Clearly, E(Xi) = E(Xj ) = 1/4. For 1 ≤ i ≤ 8, let Ai be the event that the ith card drawn is a heart. Then Xi Xj = ⎧ ⎨1 if Ai Aj occurs ⎩ 0 otherwise. Therefore, E(XiXj ) = P (Ai Aj ) = P (Aj | Ai )P (Ai ) = 3 12 1 · = . 51 4 51 Hence Cov(Xi, Xj ) = E(XiXj ) − E(Xi)E(Xj ) = 1 1 3 − =− . 51 16 272 2. Suppose that X and Y are the lifetimes of two devices each with mean 1/λ. V is the time when the first device dies; U − V is the additional lifetime of the other device, which, by the memoryless property of the exponential, is itself an exponential random variable with mean 1/λ independently of V . So, by independence of U − V and V , Cov(U − V, V ) = 0. Hence Cov(U, V ) = Cov(U − V, V ) + Cov(V, V ) = 0 + Var(V ). Note that P (V > t) = P min(X, Y ) > t = P (X > t, Y > t) = P (X > t)P (Y > t) = e−λt · e−λt = e−2λt implies that V is exponential with mean 1/(2λ) and variance 1/(4λ 2). Therefore, Cov(U, V ) = Var(V ) = 1/(4λ2). Section 10.3 1. With these answers, we have Cov(X, Y ) = E(XY ) − E(X)E(Y ) = 23 − 4 = 19, Var(X) = 13 − 4 = 9, and Var(Y ) = 40 − 4 = 36. Therefore, ρ(X, Y ) = 19 Cov(X, Y ) > 1. = σX σY 18 This is not possible, since −1 ≤ ρ(X, Y ) ≤ 1. Chapter 10 Solutions to Self-Quiz Problems 4 2. Note that n Var(X1 + X2 + · · · + Xn ) = Var(Xi ) + 2 Cov(Xi, Xj ) i=1 i<j n−1 n Cov(Xi , Xj ). = 4n + 2 i=1 j=i+1 Now ρ(Xi, Xj ) = 1 Cov(Xi , Xj ) Cov(Xi , Xj ) =− = σXi σXj 4 16 implies that Cov(X i , Xj ) = −1/4. Since n−1 i=1 n j=i+1 Cov(Xi , Xj ) has (n − 1) + (n − 2) + · · · + 1 = (n − 1)n 2 identical terms, we have Var(X1 + X2 + · · · + Xn ) = 4n + (n − 1)n − 1 n(17 − n) = . 4 4 Section 10.4 1. Due to the memoryless property of exponential random variables, it does not matter who cried last or how many times before. Let X be the time until Hannah cries next and Y be the time until Joshua cries next; X and Y are independent exponential random variables with parameters 10 and 7, respectively. By Remark 10.2, the desired probability is 7 7 = . P (Y < X) = 7 + 10 17 2. Let midnight be labeled t = 0. It should be clear that the percentage of the cars parked illegally in the time interval (0, 30) and caught is equal to the percentage of all cars parked illegally and caught. Between 0 and 30, let Y be the time a random car is illegally parked on the street; Y is a uniform random variable over (0, 30). Suppose that the car parked illegally at Y is left parked for X units of time. Let A be the event that it is caught. The desired percentage is 100 · P (A). 30 30 1 1 dy P (A | Y = y) dy = P (X > 30 − y) P (A) = 30 30 0 0 30 1 dy = 0.317. e−(1/10)(30−y) · = 30 0 Therefore, the percentage of the cars that are parked illegally and caught is 31.7%. Note that, as expected, the final answer does not depend on λ. Chapter 10 Solutions to Self-Quiz Problems 5 Section 10.5 1. The conditional probability density function of of Y , given that X = 0.97 is normal with mean μY + ρ σY 0.19 (x − μX ) = 0.53 + (0.2) (0.97 − 0.95) ≈ 0.5327 σX 0.28 and standard deviation σY 1 − ρ2 = (0.19) 1 − (0.2)2 ≈ 0.1862. Therefore, the desired probability is calculated as follows: 0.60 − 0.5327 Y − 0.5327 ≥ P (Y ≥ 0.60 | X = 0.97) = P X = 0.97 0.1862 0.1862 = P (Z ≥ 0.36) = 1 − Φ(0.36) = 1 − 0.6409 = 0.3594, where Z is a standard normal random variable and Φ is its distribution function. 2. By (10.24), f (x, y) is maximum if and only if Q(x, y) is minimum. Let z 1 = y − μY x − μX and z2 = . Then |ρ| ≤ 1 implies that σX σY Q(x, y) = z12 − 2ρz1 z2 + z22 ≥ z12 − 2|ρz1 z2 | + z22 ≥ z12 − 2|z1 z2 | + z22 = |z1 | − |z2 | 2 ≥ 0. This inequality shows that Q is minimum if Q(x, y) = 0. This happens at x = μ X and y = μY . Therefore, (μX , μY ) is the point at which the maximum of f is obtained. Chapter 10 Solutions to Self-Test Problems 6 SOLUTIONS TO SELF-TEST PROBLEMS 1. Let X be the period between two consecutive calls made to extension 1247; X is an exponential random variable with mean 1/λ. We have ∞ P N2 (X) = i | X = x λe−λx dx P (N = i) = P N2 (X) = i = = = ∞ 0 μi λ i! 0 −λx P N2 (x) = i λe ∞ 0 dx = ∞ 0 e−μx (μx)i −λx λe dx i! xi e−(λ+μ)x dx. Making the substitution (λ + μ)x = y yields ∞ μi λ 1 yi 1 μi λ ∞ −y dy = · e e−y y i dy P (N = i) = i! 0 (λ + μ)i λ+μ i! (λ + μ)i+1 0 μ i λ μi λΓ(i + 1) = = i! (λ + μ)i+1 λ+μ λ+μ i λ λ , i = 0, 1, 2, . . . . = 1− λ+μ λ+μ 2. Let X be the annual calamity loss of the business in a random year. Let Y be the annual calamity loss of the business that year paid by the insurance company. We are interested in E(Y ): E(Y ) = E E(Y | X) = = 0.4 0.25 0.4 0.25 E(Y | X = x)f (x) dx + xf (x) dx + ∞ 0.4 ∞ 0.4 E(Y | X = x)f (x) dx (0.4)f (x) dx ∞ 1 1 x· dx + (0.4) dx = 5 5 64x 0.25 0.4 64x 0.4 ≈ 0.251953 + 0.061035 = 0.312988. Therefore, the expected value of the annual loss of the business paid by the insurance company is $312,988. 3. Note that n Var(X1 + X2 + · · · + Xn ) = Var(Xi ) + 2 i=1 Cov(Xi, Xj ), i<j Chapter 10 n−1 i<j 7 n Cov(Xi, Xj ) = where Solutions to Self-Test Problems Cov(Xi, Xj ) has i=1 j=i+1 (n − 1) + (n − 2) + · · · + 1 = (n − 1)n 2 terms, and Cov(Xi , Xj ) = ρ(Xi, Xj )σXi σXj = aσ 2 . Hence Var(X1 + X2 + · · · + Xn ) ≥ 0 yields nσ 2 + (n − 1)naσ 2 ≥ 0, 1 . which implies that 1 + (n − 1)a ≥ 0 or a ≥ − n−1 4. Clearly X is exponential with parameter λ. Let Y 1 , Y2 , and Y3 be the lifetimes of the electron guns of Dan’s monitor. Then Y 1 , Y2 , and Y3 are independent exponential random variables each with parameter λ. Hence, Y = min(Y1 , Y2 , Y3 ) is exponential with parameter 3λ. This gives P (X < Y ) = 1 λ = ; λ + 3λ 4 1 . E min(X, Y ) = 4λ To find E max(X, Y ) , note that, by Remark 6.4, ∞ P max(X, Y ) > t dt E max(X, Y ) = 0 ∞ 1 − P max(X, Y ) ≤ t dt = 0 = = ∞ 1 − P (X ≤ t, Y ≤ t) dt 1 − P (X ≤ t)P (Y ≤ t) dt 0 ∞ 0 ∞ 1 − (1 − e−λt )(1 − e−3λt) dt 0 1 1 −3λt 1 −4λt ∞ 13 e e λ. + = = − e−λt − λ 3λ 4λ 12 0 = 5. Let X be the time until the seismograph placed in the New Mexico desert is replaced. Let L be its lifetime. L is an exponential random variable with parameter λ = 1/7. Let ⎧ if L ≥ 5 ⎨1 Y = ⎩ 0 if L < 5. Chapter 10 Solutions to Self-Test Problems 8 We are interested in E(X). We have E(X) = E E(X | Y ) = E(X | Y = 0)P (Y = 0) + E(X | Y = 1)P (Y = 1) = E(L | L < 5)P (L < 5) + E(X | L ≥ 5)P (L ≥ 5). Clearly, P (L < 5) = 5 0 1 −t/7 e dt = 1 − e−5/7 ≈ 0.51, 7 P (L ≥ 5) ≈ 1 − 0.51 = 0.49. E(X | L ≥ 5) = 5. To find E(L | L < 5), first we find the distribution function of L given that L < 5, we have ⎧ ⎪ 0 t<0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ P (L ≤ t) 0≤t<5 P (L ≤ t | L < 5) = ⎪ P (L < 5) ⎪ ⎪ ⎪ ⎪ ⎪ ⎩1 t ≥ 5. So ⎧ ⎪ 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ 1 − e−t/7 P (L ≤ t | L < 5) = ⎪ ⎪ 1 − e−5/7 ⎪ ⎪ ⎪ ⎪ ⎩1 t<0 0≤t<5 t ≥ 5. Let fL|L<5 be the conditional probability density function of L given L < 5. f L|L<5 is obtained by differentiating the right side of the distribution function of L given that L < 5. Thus ⎧ ⎪ (1/7)e−t/7 ⎪ ⎨ 0≤t<5 1 − e−5/7 fL|L<5 (t) = ⎪ ⎪ ⎩0 otherwise. Therefore, 1/7 E(L | L < 5) = 1 − e−5/7 = 5 0 te−t/7 dt 5 1 −t/7 − (7t + 49)e ≈ 2.205. 0 7(1 − e−5/7 ) So the probability we are interested in is E(X) = E(L | L < 5)P (L < 5) + E(X | L ≥ 5)P (L ≥ 5) = (2.205)(0.51) + (5)(0.49) ≈ 3.57. Chapter 10 Solutions to Self-Test Problems 9 y x Figure 10a Region R of Self-Test Problem #6 of Chapter 10. 6. Region R is the shaded area of Figure 10a. Since the area of R is 9π − 4π = 5π, the joint probability density function of X and Y is given by ⎧ 1 ⎪ if (x, y) ∈ R ⎨ f (x, y) = 5π ⎪ ⎩ 0 otherwise. As Figure 10b shows, P (0 < X < 1) = 2P (0 < X < 1 | Y > 0). Since P (0 < X < 1 | Y > 0) = P (0 < X < 1), X and Y are dependent. To show that X and Y are uncorrelated, we need to calculate E(X), E(Y ), and E(XY ). To calculate f (x, y) dx dy, E(X) = R we convert from Cartesian to polar coordinates. We have x = r cos θ, y = r sin θ, and dx dy = r dr dθ. So 2π 3 2π 3 1 1 2 r dr dθ = (r cos θ) · (r cos θ) r dr dθ E(X) = 5π 5π 0 0 2 2 2π 2π 19 19 sin θ cos θ dθ = = 0. = 15π 0 15π 0 Chapter 10 y Solutions to Self-Test Problems 10 x =1 x Figure 10b The shaded regions show that P (0 < X < 1 | Y > 0) = P (0 < X < 1) Similarly E(Y ) = 0. The fact that E(X) = E(Y ) = 0 must be evident since X and Y assume positive and negative values with the same probabilities. Now we calculate E(XY ). We have 2π 3 1 r dr dθ 5π 0 2 3 2π 1 3 sin θ cos θ r dr dθ = 5π 0 2 13 2π 1 13 2π sin 2θ dθ sin θ cos θ dθ = = 4π 0 4π 0 2 2π 1 1 1 13 13 − cos 2θ − + = 0. = = 8π 2 8π 2 2 0 E(XY ) = (r cos θ)(r sin θ) · Hence Cov(X, Y ) = E(XY ) − E(X)E(Y ) = 0 − 0 · 0 = 0. This shows that X and Y are uncorrelated. 7. Let N (t) be the number of claims received by the insurance company at or prior to time t. Let Λ be the average number of claims during a one month period. We are given that N (t) : t ≥ 0 is a Poisson process with parameter Λ, where Λ is a gamma Chapter 10 Solutions to Self-Test Problems 11 random variable with parameters n and λ. To calculate the distribution of N (12), the desired random variable, we will condition on Λ. ∞ λe−λx (λx)n−1 dx P N (12) = i | Λ = x P N (12) = i = (n − 1)! 0 ∞ −12x e (12x)i λe−λx (λx)n−1 · dx = i! (n − 1)! 0 ∞ 12iλn e−(λ+12)x xn+i−1 dx Now let y = (λ + 12)x. = i! (n − i)! 0 ∞ 12i λn 1 = · e−y y n+i−1 dy i! (n − 1)! (λ + 12)n+i 0 1 12i λn · Γ(n + i) i! (n − 1)! (λ + 12)n+i λ n (n + i − 1)! 12 i = = λ + 12 λ + 12 i! (n − 1)! n 12 i n+i−1 λ , = λ + 12 λ + 12 n−1 For i ≥ 0, letting j = n + i, we have shown that 12 j−n j−1 λ n , P n+N (12) = j = λ + 12 n − 1 λ + 12 i = 0, 1, 2, . . .. j = n, n+1, n+2, . . .. This shows that n + N (12) is negative binomial with parameter n and λ (λ + 12). 8. For i = 1, 2, . . ., 10, j = i + 1, i + 2, . . ., 11, and k = j + 1, j + 2, . . . , 12, let Xijk = k if the ith, the jth, and the kth movies withdrawn are Marlon Brando 10 11 12 movies. Let Xijk = 0, otherwise. Then X = Xijk is the number i=1 j=i+1 k=j+1 of movies Professor Jackson withdraws until all Brando movies are pulled out. The desired quantity is 10 11 12 E(X) = E 10 Xijk = i=1 j=i+1 k=j+1 10 11 12 = i=1 j=i+1 k=j+1 k· 11 12 E(Xijk ) i=1 j=i+1 k=j+1 10 11 12 1 1 = k. 12 220 i=1 j=i+1 k=j+1 3 Now 12 k=j+1 j 12 k= k− k=1 k= k=1 1 12 × 13 j(j + 1) − = (156 − j 2 − j). 2 2 2 Chapter 10 Solutions to Self-Test Problems 12 So 11 12 j=i+1 k=j+1 11 1 k= 2 = 1 2 (156 − j 2 − j) j=i+1 11 (156 − j 2 − j) − j=1 i (156 − j 2 − j) j=1 = 1 11 × 12 × 23 11 × 12 1716 − − − 156i 2 6 2 i(i + 1)(2i + 1) i(i + 1) + + 6 2 = 572 − 78i + i(i + 1)(i + 2) . 6 Therefore, 1 E(X) = 220 10 i=1 572 − 78i + i(i + 1)(i + 2) 6 1 78 10 × 11 · + = 26 − 220 2 220 = 26 − 10 i=1 i(i + 1)(i + 2) 6 1 39 39 + · 4290 = = 9.75. 2 1320 4 9. Let X1 = 1 for counting the first person that Amber will meet. Let X 2 be the number of people Amber will meet until she finds a person whose birthday is different from the birthday of the first person she met. Clearly, X 2 is a geometric random variable with probability of success p given by p = 364/365. Let X 3 be the number of people Amber will meet until she finds a person whose birthday is different from the first two birthdays that she has already recorded. Then X 3 is geometric with p = 363/365. In general, for 2 ≤ i ≤ 365, let Xi be the number of people Amber will meet until she finds a person whose birthday is different from the first i − 1 birthdays she has already recorded. Then Xi is a geometric random variable with p= 366 − i 365 − (i − 1) = . 365 365 Clearly, X, the number of people that Amber will need to meet in order to achieve her goal, satisfies the following equation. X = X1 + X2 + X3 + · · · + X365. Since the expected value of a geometric random variable with parameter p is 1/p, we Chapter 10 Solutions to Self-Test Problems 13 have E(X) = E(X1) + E(X2) + E(X3 ) + · · · + E(X365) = 1+ 365 365 365 + +···+ 364 363 1 364 = 1 + 365 i=1 1 ≈ 1 + 365(6.47574252996) ≈ 2364.65. i Now, since the variance of a geometric random variable with parameter p is and since X1 , X2 , . . . , X365 are independent random variables, 365 365 Var(Xi ) = Var(X) = i=1 = Var(Xi ) i=2 366 − i 365 365(i − 1) 365 = ≈ 216, 417.19. 2 (366 − i)2 366 − i i=2 365 365 1 − i=2 Therefore, σX = 1−p , p2 Var(X) ≈ 465.20. 10. The set of possible outcomes of this experiment is N N N N, 6NN N, N 6N N, NN 6N, NN N 6, 66NN, 6N6N, 6NN 6, N66N, N 6N 6, N N 66, 666N, 66N 6, 6N 66, N666, 6666 . The probability of obtaining N N N N is (5/6) 4 = 625/1296. The probability of each of the possible outcomes 6N N N , N 6N N , N N 6N , and N N N 6 is (1/6)(5/6) 3 = 125/1296. The probability of each of the possible outcomes 66N N , 6N 6N , 6N N 6, N 66N , N 6N 6, and N N 66 is (1/6)2(5/6)2 = 25/1296. The probability of each of the possible outcomes 666N , 66N 6, 6N 66, N 666 is (1/6) 3(5/6) = 5/1296. Finally, the probability of obtaining 6666 is (1/6) 4 = 1/1296. Based on these probabilities, for example, P (X = 2, Y = 2) = P {66N N, N 66N, NN66} = 25 25 75 25 + + = . 1296 1296 1296 1296 Similar calculations yield the following table, which represents p(x, y), the joint probability mass function of X and Y , p X (x), the marginal probability mass function of X, and pY (y), the marginal probability mass function of Y . Chapter 10 14 Solutions to Self-Test Problems y x 0 1 2 3 4 p X (x) 0 625/1296 0 0 0 0 625/1296 1 0 500/1296 0 0 0 500/1296 2 0 75/1296 75/1296 0 0 150/1296 3 0 0 10/1296 10/1296 0 20/1296 4 0 0 0 0 1/1296 1/1296 pY (y) 625/1296 575/1296 85/1296 10/1296 1/1296 To calculate the correlation coefficient of X and Y , among other quantities, we need to calculate Cov(X, Y ) and hence E(XY ). So first we calculate the probability mass function of XY , which is represented in the following table. i 0 1 2 3 4 6 8 9 12 16 P (XY = i) 625 1296 500 1296 75 1296 0 75 1296 10 1296 0 10 1296 0 1 1296 Now we will calculate all the quantities that we need to find ρ(X, Y ). We have E(X) = 0 · 500 150 20 1 864 2 625 +1· +2· +3· +4· = = , 1296 1296 1296 1296 1296 1296 3 1296 625 500 150 20 1 + 12 · + 22 · + 32 · + 42 · = = 1, 1296 1296 1296 1296 1296 1296 2 2 2 5 = , Var(X) = E(X 2) − E(X) = 1 − 3 9 √ 5 . σX = 3 E(X 2) = 02 · E(Y ) = 0 · 575 85 10 1 779 625 +1· +2· +3· +4· = , 1296 1296 1296 1296 1296 1296 1021 625 575 85 10 1 + 12 · + 22 · + 32 · + 42 · = , 1296 1296 1296 1296 1296 1296 2 779 2 1021 716, 375 − , = Var(Y ) = E(Y 2 ) − E(Y ) = 1296 1296 1, 679, 616 E(Y 2 ) = 02 · Chapter 10 σY = E(XY ) = 0 · 15 √ 6480 28, 655 716, 375 = . 1, 679, 616 1, 679, 616 500 75 75 625 +1· +2· +3·0+4· 1296 1296 1296 1296 +6· 10 1 31 10 +8·0+9· + 12 · 0 + 16 · = . 1296 1296 1296 36 Cov(X, Y ) = E(XY ) − E(X)E(Y ) = ρ(X, Y ) = Solutions to Self-Test Problems 895 31 2 779 − · = , 36 3 1296 1944 895/1944 Cov(X, Y ) = √ √ ≈ 0.946. σX σY 5 6480 28, 655 3 1, 679, 616 As expected, this shows that there is almost a linear relation, hence a strong positive correlation, between X and Y .