1 Chapter5. Joint random variables Problem PP143 1. When you roll one more time you receive π + π when you roll k (probability = 1⁄2π) with the die provided π + π ≤ 2π, otherwise you receive nothing. Hence, πΈπΆ (π ) = 2. 1 4π 1 2π ∗ [(π + 1) + (π + 2) + … + 2π]; ∗ (2π − π ) ∗ (2π + π + 1) = π or π 2 + (4π + 1) ∗ π − 4π2 − 2π = 0 1 Solving the quadratic equation: π π = 2 ∗ (−4π − 1 + √32π2 + 16π + 1) (the second root is negative). Notice that continuing to roll is better than stopping when the left hand side of the quadratic equation is smaller than 0. Also the quadratic function has a convex shape, hence π ∗ = ⌊π π ⌋. Problem PP144 1. Straightforward; 2. Let π = π1 + π2 and π = π1. The joint pdf of S and T is π(π , π‘) = 2 ∗ (π − π‘) , 0 ≤ π‘ ≤ 1 πππ π‘ ≤ π ≤ π‘ + 1 The marginal pdf of S: π π(π ) = ∫ 2 ∗ (π − π‘) ∗ ππ‘ = π 2 , 0 ≤ π ≤ 1 0 1 = ∫π −1 2 ∗ (π − π‘) ∗ ππ‘ = π ∗ (2 − π ) , 1 ≤ π ≤ 2 1 2 4. πΈ(π) = ∫0 π 3 ∗ ππ + ∫1 π 2 ∗ (2 − π ) ∗ ππ = 7⁄6 = 1.16667 or: πΈ(π) = πΈ(π1 ) + πΈ(π2 ) = 7⁄6 5. Straightforward 2 6. Let M denote the median. Using the result in Step 5: − π3 ⁄3 + π2 − 1⁄3 = 1⁄2. Use the Solver to find π = 1.168; Problem PP145 1 1−π₯1 1. ∫0 ∫0 π ∗ π₯1 ∗ π₯2 ∗ ππ₯2 ∗ ππ₯1 = 1 βΉ π = 24 1−π₯ 2. π(π₯1 ) = ∫0 1 24 ∗ π₯1 ∗ π₯2 ∗ ππ₯2 = 12 ∗ π₯1 ∗ (1 − π₯1 )2 , 0 ≤ π₯1 ≤ 1: beta pdf with parameters πΌ = 2 and π½ = 3. Because of symmetry the pdf of π2 has the same pdf; 3. πΈ(π1 ) = πΈ(π2 ) = πΌ πΌπ½ πΌ+π½ = 2⁄5 , π£ππ(π1 ) = π£ππ(π2 ) = (πΌ+π½)2 ∗(πΌ+π½+1) = 1⁄25; 4. Let π = π1 + π2 and π = π2. Then π1 = π − π and π2 = π. The J(acobian) is π½ = 1. π(π , π‘) = 24 ∗ π‘ ∗ (π − π‘) , 0 ≤ π‘ ≤ π ≤ 1 π π(π ) = ∫ 24 ∗ π‘ ∗ (π − π‘) ∗ ππ‘ = 4 ∗ π 3 , 0 ≤ π ≤ 1 0 1 5. πΈ(π) = ∫0 π ∗ 4 ∗ π 3 ∗ ππ = 4⁄5 = 0.8 2⁄75 1 1−π₯1 π₯1 6. πΈ(π1⁄π2 ) = ∫0 ∫0 π₯2 24∗π₯1 ∗π₯2 7. π(π₯2 |π₯1 ) = 12∗π₯ 2 1 ∗(1−π₯1 ) π£ππ(π) = πΈ(π 2 ) − [πΈ(π)]2 = 2⁄3 − 16⁄25 = ∗ 24 ∗ π₯1 ∗ π₯2 ∗ ππ₯2 ∗ ππ₯1 = 2 π₯ = 2 ∗ (1−π₯2 2 1) , 0 ≤ π₯2 ≤ 1 − π₯1 Problem PP146 π π(π₯, π) = ( ) ∗ π π₯ ∗ (1 − π)π−π₯ ∗ 1 , 0 ≤ π ≤ 1 , π₯ π₯ = 0, 1, 2, … , π 1 π π(π₯) = ( ) ∗ ∫ π π₯ ∗ (1 − π)π−π₯ ∗ ππ = π₯ 0 1 π Γ(π₯ + 1) ∗ Γ(π + π₯ + 1) Γ(π + 2) ( )∗ ∗∫ ∗ π π₯ ∗ (1 − π)π−π₯ ∗ ππ = π₯ Γ(π + 2) Γ(π₯ + 1) ∗ Γ(π + π₯ + 1) 0 3 = (ππ₯) ∗ Γ(π₯+1)∗Γ(π+π₯+1) Γ(π+2) 1 = π+1 , π₯ = 0, 1, 2, … , π (integrand is a beta pdf) Problem PP148 1. π(π1 ) = 1 , 1 ≤ π1 ≤ 2 and likewise for π2 . πΉ(π1 ) = π(πΏ1 ≤ π1 ) = π1 − 1 , 1 ≤ π1 ≤ 2 π(πΏ ≤ π) = πΉ(π) = π(πΏ1 ≤ π β πΏ2 ≤ π) = (π − 1)2 , 1 ≤ π ≤ 2 π(π) = 2 ∗ (π − 1) , 1 ≤ π ≤ 2 2 2. πΈ(πΏ) = 2 ∗ ∫1 π ∗ (π − 1) ∗ ππ = 5⁄3 2 2) πΈ(πΏ = 2 ∗ ∫ π 2 ∗ (π − 1) ∗ ππ = 17⁄6 1 π£ππ(πΏ) = πΈ(πΏ2 ) − [πΈ(πΏ)]2 = 1⁄18 3. πΏπ = πΏ1 + πΏ2 , 2 ≤ πΏπ ≤ 4 Let ππ = π1 + π2 and ππ‘ = π1 . Then π1 = ππ‘ and π2 = ππ‘ − ππ , π½(πππππππ) = 1 π(ππ‘ , ππ ) = 1 , 1 ≤ ππ‘ ≤ 2 , ππ‘ + 1 ≤ ππ ≤ ππ‘ + 2 ππ −1 π(ππ ) = ∫ = πππ‘ = ππ − 2 , 2 ≤ ππ ≤ 3 1 2 ∫π −2 πππ‘ π = 4 − ππ , 3 ≤ ππ ≤ 4 4. πΈ(πΏπ ) = πΈ(πΏ1 + πΏ2 ) = πΈ(πΏ1 ) + πΈ(πΏ2 ) = 1 π£ππ(πΏπ ) = π£ππ(πΏ1 ) + π£ππ(πΏ2 ) = 1⁄6 Problem PP149 1. πΈ(2 ∗ πΏ + 2 ∗ π) = 2 ∗ πΈ(πΏ) + 2 ∗ πΈ(π) = 2 ∗ 100 + 2 ∗ 70 = 340 π£ππ(2 ∗ πΏ + 2 ∗ π) = 4 ∗ π£ππ(πΏ) + 4 ∗ π£ππ(π) = 4 ∗ 100 + 4 ∗ 25 = 500 C is a linear combination of Normally distributed random variables, hence also normally distributed; 4 2. π(300 ≤ πΆ ≤ 350) = πππ π. π·πΌππ(350,340, πππ π(500), 1) − πππ π. π·πΌππ(300,340, πππ π(500), 1) = 0.6358; 3. πΈ(π΄) = πΈ(πΏ ∗ π) = πΈ(πΏ) ∗ πΈ(π) = 100 ∗ 70 = 7000 π2 π£ππ(π΄) = π£ππ(πΏ ∗ π) = πΈ(πΏ2 ∗ π 2 ) − πΈ 2 (πΏ ∗ π) = πΈ(πΏ2 ) ∗ πΈ(π 2 ) − 70002 = [π£ππ(πΏ) + πΈ 2 (πΏ)] ∗ [π£ππ(π) + πΈ 2 (π)] − 70002 = 742500 π4 Problem PP150 1. Let πΉ1 (πΉ2 ) be the end of operation 1 (operation 2). Then πΉ1 − πΉ2 is normally distributed with πΈ (πΉ1 − πΉ2 ) = −1 and π£ππ(πΉ1 − πΉ2 ) = 4. π(πΉ1 ≤ πΉ2 ) = π(πΉ1 − πΉ2 ≤ 0) = πππ π. π·πΌππ(0, −1,2,1) = 0.6914; 2. π(−1 ≤ πΉ1 − πΉ2 ≤ 0) = πππ π. π·πΌππ(0, −1,2,1) − πππ π. π·πΌππ(−1, −1,2,1) = 0.1915; 3. Let X be the time that operation 2 starts after the start of operation 1 (in minutes). πΉ1 − πΉ2 ∼ π(5 − π, 2). The probability that part 1 will be available before part 2 but by no more than 3 minutes: π(−3 ≤ πΉ1 − πΉ2 ≤ 0) = πππ π. π·πΌππ(0,5 − π, 2,1) − πππ π. π·πΌππ(−3,5 − π, 2,1) which should be maximized. Problem PP151 1. Let ππ be the weight of sweet i. Then ππ ∼ π(5 , √0.5 ). The sum ∑ππ=1 ππ of the weight of n sweets is normal with expected value 5 ∗ π and variance 0.5 ∗ π. The number of sweets required is the smallest integer such that π ∑ππ=1 ππ − 5 ∗ π 200 − 5 ∗ π π (∑ ππ > 200) = π ( > ) ≥ 0.999 π=1 √0.5 ∗ π √0.5 ∗ π n is the smallest integer larger than the solution of 200 − 5 ∗ π √0.5 ∗ π = −3.09 or 5 ∗ π − 3.09 ∗ √0.5 ∗ π − 200 = 0 βΉ π = 45 5 Problem PP152 1. Let A (B) denote the arrival time of A (B). Let DTA (DTB) denote the journey time of A (B). πΈ(π΄) = 4 ππ πΈ(π·ππ΅) = 5 + 1⁄2 + 2 + 1⁄2 + 2.5 = 10.5 βππ βΉ πΈ(π΅) = 4.30 ππ 2. DTB is normally distributed with expected value 630 minutes and variance 900 + 100 + 225 = 1225 ππππ’π‘ππ 2 . π(π΄ < 5 ππ) = π(π·ππ΄ < 660 πππ) = 0.84 π(π΅ < 5 ππ) = π(π·ππ΅ < 660 πππ) = 0.804 3. π(π΅ πππππ£ππ ππππππ π΄) = π(π·ππ΅ − π·ππ΄ < 0) = π (π§ < 0−30 ) = 0.333 √1225+3600 Problem PP153 1. Let ππ be the length of step k with pmf π(ππ = 1) = 0.5 and π(ππ = −0.8) = 0.5 πΈ(ππ ) = 0.1 π , π£ππ(ππ ) = 0.81 π2 The sum ∑900 π=1 ππ is approximately normally distributed with expected value 90 and variance 729. Hence, approximately, 900 π (∑ ππ ≥ 100) = π (π ≥ π=1 100 − 90 √729 ) = 0.355 2. The pdf of ππ is uniform: π(ππ ) = 1⁄1.8 , −0.8 ≤ π₯π ≤ 1 πΈ(ππ ) = 0.1 π , π£ππ(ππ ) = 0.27 π2 900 π (∑ ππ ≥ 100) = π (π ≥ π=1 3. The pmf of ππ is π(π₯π ) = 1⁄4 , π₯π = 1 = 1⁄2 , π₯π = 0 = 1⁄4 , π₯π = −1 100 − 90 √243 ) = 0.2606 6 πΈ(ππ ) = 0 π , π£ππ(ππ ) = 0.5 π2 900 π (∑ ππ ≥ 10) = π (π ≥ 0 − 10 π=1 √450 ) = 0.3187 Problem PP156 1. We solve the problem for general parameters π1 and π2 . Let π¦ = π₯1 ⁄π₯2 and π‘ = π₯2 . Then π₯1 = π¦ ∗ π‘ and π₯2 = π‘. The Jacobian of the transformation: π½ = π‘. π(π¦, π‘) = π1 ∗ π2 ∗ π −π1 ∗π¦∗π‘ ∗ π −π2 ∗π‘ ∗ π‘ , π¦, π‘ > 0 ∞ π(π¦) = π1 ∗ π2 ∗ ∫ π‘ ∗ π −π‘∗(π2 +π¦∗π1 ) ∗ ππ‘ 0 Notice that the integrand is, except for the constant (π2 + π¦ ∗ π1 )2 , a gamma density with parameters πΌ = 2 and πΎ = π2 + π¦ ∗ π1 . Hence, π(π¦) = π1 ∗ π2 , π¦>0 (π2 + π¦ ∗ π1 )2 2. ∞ 3. π(π > π) = π1 ∗ π2 ∗ ∫π 1 (π2 +π¦∗π1 )2 ∗ ππ¦ = π π2 2 +π∗π1 . For π1 = 4, π2 = 2 and π = 2: π(π > 2) = 0.2 Problem PP157 1. π = πππ₯(π, π) where X (Y) is the time to finish project 1 (project 2). π₯ 1 1 πΉπ (π₯) = ∫10 10 ∗ ππ₯ = 10 ∗ (π₯ − 10) , 10 < π₯ < 20. Similarly for Y. π(π ≤ π‘) = π(π ≤ π‘ ∩ π ≤ π‘) = 1 ∗ (π‘ − 10)2 , 10 < π‘ ≤ 20 150 1 = 1 ∗ 15 ∗ (π‘ − 10) , 20 < π‘ ≤ 25 7 π(π‘) = 1 ∗ (π‘ − 10) , 10 < π‘ ≤ 20 75 1 = 15 , 20 < π‘ ≤ 25; 1 43 2. π(π > 18) = 1 − 150 ∗ (18 − 10)2 = 75 πΈ(ππππ) = 75000 ∗ 43 = 43000 75 πΈ(ππππππ‘) = 50000 − 43000 = € 7000 Problem PP159 1. Let π€ = π₯ 2 ∗ y and π£ = π¦. Then π₯ = √π€ ⁄π£ and π¦ = π£. The J(acobian) is 1 π½ = 2 ∗ (π£ ∗ π€)−1⁄2. π(π£, π€) = 6 ∗ [1 − (π€ ⁄π£)1⁄2 ] , 0 ≤ π€ ≤ π£ ≤ 1 1 π(π€) = 6 ∗ ∫ [1 − (π€ ⁄π£)1⁄2 ] ∗ ππ£ = 6 ∗ (1 + π€ − 2 ∗ √π€) , 0 ≤ π€ ≤ 1 π€ 1 2. πΈ(π) = 6 ∗ ∫0 π€ ∗ (1 + π€ − 2 ∗ √π€) ∗ ππ€ = 1⁄5 Problem PP160 1. π(π¦|π₯) = π(π₯, π¦) = π(π¦) = 1 √2π 1 2 ∗ π −2∗(π¦−2∗π₯) , −∞ < π¦ < +∞ 1 2 1 1 1 2 2 2 ∗ π −2∗[π₯ +(π¦−2∗π₯) ] = ∗ π −2∗(5∗π₯ +π¦ −4∗π₯∗π¦) , 2π 2π −∞ < π₯, π¦ < +∞ +∞ 5 1 2 1 2 4 ∗ π −2∗π¦ ∗ ∫ π −2∗(π₯ −5∗π₯∗π¦) ∗ ππ₯ 2π −∞ +∞ 5 1 2 2 2 1 2 4 2 4 = ∗ π −2∗π¦ ∗ π 5∗π¦ ∗ ∫ π − 2∗(π₯ +25π¦ −5∗π₯∗π¦) ∗ ππ₯ 2π −∞ 2 +∞ 1 2 1 2 1 1 1 − ∗(π₯− ∗π¦) − ∗π¦ 5 = ∗ π 10 ∗ √1⁄5 ∗ ∫ ∗ ∗ π 2∗(1⁄5) ∗ ππ₯ √2π −∞ √2π √1⁄5 8 = 1 1 √2π∗√5 2 ∗ π −2∗5∗π¦ : normal with expected value 0 and variance 5; 2. Let π£ = π₯ + π¦ and π€ = π₯. Then π¦ = π£ − π€ and π₯ = π€. The Jacobian of the transformation is 1. π(π£, π€) = π(π£) = = = 1 1 2 2 ∗ π − 2∗[π€ +(π£−3π€) ] , 2π −∞ < π£, π€ < +∞ +∞ 10 1 2 1 2 6 ∗ π −2∗π£ ∗ ∫ π − 2 ∗(π€ −10∗π£∗π€) ∗ ππ€ 2π −∞ 1 ∗π √2π 1 2 − ∗π£ 2 1 √2π∗√10 ∗π 1 9 20 1 ∗ √10 ∗ +∞ 1 ∫−∞ √2π ∗ 1 √1⁄10 1 ∗ 3 − ∗(π€− ∗π£) 10 π 2⁄10 2 ∗ ππ€ 2 ∗ π −2∗10∗π£ : normal with expected value 0 and variance 10 Problem PP162 1. Call your arrival time 0, X the arrival time of the first arriving bus and Y the arrival time of the first arriving cab. Then π(π₯) = 1 , 0 ≤ π₯ ≤ 1 and π(π¦) = π − π¦ , π¦ > 0 π = πππ(π, π) π(π > π€) = 0 , π€ > 1 = π(π > π€ ∩ π > π€) = (1 − π€) ∗ π − π€ πΉ(π€) = 1 − (1 − π€) ∗ π − π€ π(π€) = (2 − π€) ∗ π − π€ , 0 ≤ π€ ≤ 1 1 πΈ(π) = ∫ π€ ∗ (2 − π€) ∗ π − π€ ∗ ππ€ = π −1 = 0.368 0 1 π₯ 2. π(π‘ππ₯π ππππππ ππ’π ) = π(π < π) = ∫0 ππ₯ ∗ ∫0 π − π¦ ∗ ππ¦ = π − 1 = 0.368 9 Problem PP163 1 Γ(πΌ+π½) 1. ππ (π) = ∫0 (ππ) ∗ π π ∗ (1 − π)π−π ∗ Γ(πΌ)∗Γ(π½) ∗ π πΌ−1 ∗ (1 − π)π½−1 ∗ ππ π! 1 Γ(πΌ+π½) = π !∗(π−π) ! ∗ Γ(πΌ)∗Γ(π½) ∗ ∫0 π πΌ+π−1 ∗ (1 − π)π+π½−π−1 ∗ ππ π! Γ(πΌ+π½) = π !∗(π−π) ! ∗ Γ(πΌ)∗Γ(π½) ∗ Γ(πΌ+π)∗Γ(π+π½−π) Γ(πΌ+π½+π) Problem PP166 1. π(π₯, π¦) = 2⁄π₯ , 0 < π¦ < π₯ 2 , 0 < π₯ < 1 1 Let π£ = π₯ 2 − π¦ and π€ = π¦. Then π₯ = (π£ + π€)1⁄2 and π¦ = π€. The J(acobian): π½ = 2 ∗ (π£ + π€)− 1⁄2 1 , π£ > 0 ,π€ > 0 ,0 ≤ π£ + π€ ≤ 1 π£+π€ π(π£, π€) = 1−π£ π(π£) = ∫ 0 1 ∗ ππ€ = −ππ(π£) , 0 < π£ < 1 π£+π€ 1 πΈ(π) = ∫ π£ ∗ ln(π£) ∗ ππ£ = 1⁄4 0 1 2. Let π£ = π₯ 2 + π¦ and π€ = π¦. Then π₯ = (π£ − π€)1⁄2 and π¦ = π€. The J(acobian): π½ = 2 ∗ (π£ − π€)− 1⁄2 1 , π£ > 2 ∗ π€ ,π£ ≤ π€ + 1 ,0 ≤ π€ ≤ 1 ,0 ≤ π£ ≤ 2 π£−π€ π(π£, π€) = π£⁄2 π(π£) = ∫ 0 π£⁄2 =∫ π£−1 2 1 ∗ ππ€ = ππ(2) , 0 ≤ π£ ≤ 1 π£−π€ 1 ∗ ππ€ = −ππ(π£⁄2) = ππ(2) − ππ(π£) , 1 ≤ π£ ≤ 2 π£−π€ 2 πΈ(π) = ∫ π£ ∗ ππ(2) ∗ ππ£ − ∫ π£ ∗ ππ(π£) ∗ ππ£ = 3⁄4 0 1 10 Problem PP167 ∞ 1. π(π₯) = π2 ∗ ∫π₯ π − π∗π¦ ∗ ππ¦ = π ∗ π − π∗π₯ , π₯ > 0 ; π¦ 2. π(π¦) = π2 ∗ ∫0 π − π∗π¦ ∗ ππ¦ = π2 ∗ π¦ ∗ π − π∗π¦ , π¦ > 0 ; 3. π(π¦|π₯) = π2 ∗π −π∗π¦ π∗π − π∗π₯ = π ∗ π − π∗(π¦−π₯) , π¦ > π₯ π¦ πΉ(π¦|π₯) = π ∗ π π∗π₯ ∗ ∫ π − π∗π§ ∗ ππ§ = 1 − π − π∗(π¦−π₯) , π¦ > π₯ π₯ To generate a random value of Y given π = π₯, set 1 − π − π∗(π¦−π₯) = π where R is a random number. Solving for y/ 1 π¦ = π₯ − ∗ ππ(1 − π ) π 4. Let π£ = π₯⁄π¦ and π€ = π¦. Then π₯ = π£ ∗ π€ and π¦ = π€. The J(acobian) is π½ = π€. π(π£, π€) = π2 ∗ π − π∗π€ ∗ π€ , π€ > 0 , 0 < π£ < 1 ∞ 2 π(π£) = π ∗ ∫ π€ ∗ π − π∗π€ ∗ ππ€ = 1 , 0 < π£ < 1 0 5. πΈ(π) = 1⁄2 6. Let π€ = π¦⁄π₯ and π’ = π₯. Then π¦ = π’ ∗ π€ and π₯ = π’. The J(acobian) is π½ = π’. π(π€, π’) = π2 ∗ π − π∗π€∗π’ ∗ π’ , π€ > 1 , π’ > 0 ∞ 2 π(π€) = π ∗ ∫ π’ ∗ π − π∗π’∗π€ ∗ ππ’ = 1⁄π€ 2 , π€ > 1 0 ∞ 1 7. πΈ(π) = ∫1 π€ ∗ π€ 2 ∗ ππ€ = +∞ 11 Problem PP168 1. Let π΅π denote the number of balls in box i. Then ππ = 0 π€βππ π΅π = 0 ππ = π΅π − 1 π€βππ π΅π > 0 π πΈ(ππ ) = ∑ π (π − 1) ∗ π(π΅π = π) = ∑ π=1 π π ∗ π(π΅π = π) − ∑ π=1 π=1 π(π΅π = π) = π ∗ ππ − [1 − π(π΅π = 0)] = π ∗ ππ − 1 + (1 − ππ )π as π΅π is binomially distributed with parameters n and ππ . π πΈ (∑ π ππ ) = ∑ π=1 π πΈ(ππ ) = π − π + ∑ π=1 (1 − ππ )π π=1