Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Problem Set #5 Econ 103 Part I – Problems from the Textbook Chapter 4: 1, 3, 5, 7, 9, 11, 13, 15, 25, 27, 29 Chapter 5: 1, 3, 5, 9, 11, 13, 17 Part II – Additional Problems 1. Suppose X is a random variable with support {−1, 0, 1} where p(−1) = q and p(1) = p. (a) What is p(0)? Solution: By the complement rule p(0) = 1 − p − q. (b) Calculate the CDF, F (x0 ), of X. Solution: 0, x0 < −1 q, −1 ≤ x0 < 0 F (x0 ) = 1 − p, 0 ≤ x0 < 1 1, x ≥ 1 0 (c) Calculate E[X]. Solution: E[X] = −1 · q + 0 · (1 − p − q) + p · 1 = p − q (d) What relationship must hold between p and q to ensure E[X] = 0? Solution: p = q 2. Fill in the missing details from class to calculate the variance of a Bernoulli Random Variable directly, that is without using the shortcut formula. 1 Solution: X σ 2 = V ar(X) = (x − µ)2 p(x) x∈{0,1} = X (x − p)2 p(x) x∈{0,1} = (0 − p)2 (1 − p) + (1 − p)2 p = p2 (1 − p) + (1 − p)2 p = p2 − p3 + p − 2p2 + p3 = p − p2 = p(1 − p) 3. Prove that the Bernoulli Random Variable is a special case of the Binomial Random variable for which n = 1. (Hint: compare pmfs.) Solution: The pmf for a Binomial(n, p) random variable is n x p(x) = p (1 − p)n−x x with support {0, 1, 2 . . . , n}. Setting n = 1 gives, 1 x p(x) = p(x) = p (1 − p)1−x x with support {0, 1}. Plugging in each realization in the support, and recalling that 0! = 1, we have 1! p0 (1 − p)1−0 = 1 − p p(0) = 0!(1 − 0)! and p(1) = 1! p1 (1 − p)0 = p 1!(1 − 1)! which is exactly how we defined the Bernoulli Random Variable. 4. Suppose that X is a random variable with support {1, 2} and Y is a random variable with support {0, 1} where X and Y have the following joint distribution: pXY (1, 0) = 0.20, pXY (1, 1) = 0.30 pXY (2, 0) = 0.25, pXY (2, 1) = 0.25 Page 2 (a) Express the joint distribution in a 2 × 2 table. Solution: X Y 0 1 1 0.20 0.30 2 0.25 0.25 (b) Using the table, calculate the marginal probability distributions of X and Y . Solution: pX (1) = pXY (1, 0) + pXY (1, 1) = 0.20 + 0.30 = 0.50 pX (2) = pXY (2, 0) + pXY (2, 1) = 0.25 + 0.25 = 0.50 pY (0) = pXY (1, 0) + pXY (2, 0) = 0.20 + 0.25 = 0.45 pY (1) = pXY (1, 1) + pXY (2, 1) = 0.30 + 0.25 = 0.55 (c) Calculate the conditional probability distribution of Y |X = 1 and Y |X = 2. Solution: The distribution of Y |X = 1 is P (Y = 0|X = 1) = pXY (1, 0) 0.2 = = 0.4 pX (1) 0.5 P (Y = 1|X = 1) = pXY (1, 1) 0.3 = = 0.6 pX (1) 0.5 while the distribution of Y |X = 2 is P (Y = 0|X = 2) = pXY (2, 0) 0.25 = = 0.5 pX (2) 0.5 P (Y = 1|X = 2) = pXY (2, 1) 0.25 = = 0.5 pX (2) 0.5 (d) Calculate E[Y |X]. Page 3 Solution: E[Y |X = 1] = 0 × 0.4 + 1 × 0.6 = 0.6 E[Y |X = 2] = 0 × 0.5 + 1 × 0.5 = 0.5 Hence, E[Y |X] = 0.6 with probability 0.5 0.5 with probability 0.5 since pX (1) = 0.5 and pX (2) = 0.5. (e) What is E[E[Y |X]]? Solution: E[E[Y |X]] = 0.5 × 0.6 + 0.5 × 0.5 = 0.3 + 0.25 = 0.55. Note that this equals the expectation of Y calculated from its marginal distribution, since E[Y ] = 0 × 0.45 + 1 × 0.55. This illustrates the so-called “Law of Iterated Expectations.” (f) Calculate the covariance between X and Y using the shortcut formula. Solution: First, from the marginal distributions, E[X] = 1 · 0.5 + 2 · 0.5 = 1.5 and E[Y ] = 0 · 0.45 + 1 · 0.55 = 0.55. Hence E[X]E[Y ] = 1.5 · 0.55 = 0.825. Second, E[XY ] = (0 · 1) · 0.2 + (0 · 2) · 0.25 + (1 · 1) · 0.3 + (1 · 2)0.25 = 0.3 + 0.5 = 0.8 Finally Cov(X, Y ) = E[XY ] − E[X]E[Y ] = 0.8 − 0.825 = −0.025 5. Let X and Y be discrete random variables and a, b, c, d be constants. Prove the following: (a) Cov(a + bX, c + dY ) = bdCov(X, Y ) Solution: Let µX = E[X] and µY = E[Y ]. By the linearity of expectation, E[a + bX] = a + bµX E[c + dY ] = c + dµY Thus, we have (a + bx) − E[a + bX] = b(x − µX ) (c + dy) − E[c + dY ] = d(y − µY ) Page 4 Substituting these into the formula for the covariance between two discrete random variables, XX Cov(a + bX, c + dY ) = [b(x − µX )] [d(y − µY )] p(x, y) x = bd y XX x (x − µX )(y − µY )p(x, y) y = bdCov(X, Y ) (b) Corr(a + bX, c + dY ) = Corr(X, Y ) provided that b, c are positive. Solution: Corr(a + bX, c + dY ) = p = p = p Cov(a + bX, c + dY ) V ar(a + bX)V ar(c + dY ) bdCov(X, Y ) b2 V ar(X)d2 V ar(Y ) Cov(X, Y ) V ar(X)V ar(Y ) = Corr(X, Y ) 6. Fill in the missing steps from lecture to prove the shortcut formula for covariance: Cov(X, Y ) = E[XY ] − E[X]E[Y ] Solution: By the Linearity of Expectation, Cov(X, Y ) = E[(X − µX )(Y − µY )] = E[XY − µX Y − µY X + µX µY ] = E[XY ] − µx E[Y ] − µY E[X] + µX µY = E[XY ] − µX µY − µY µX + µX µY = E[XY ] − µX µY = E[XY ] − E[X]E[Y ] Page 5 7. Let X1 be a random variable denoting the returns of stock 1, and X2 be a random variable denoting the returns of stock 2. Accordingly let µ1 = E[X1 ], µ2 = E[X2 ], σ12 = V ar(X1 ), σ22 = V ar(X2 ) and ρ = Corr(X1 , X2 ). A portfolio, Π, is a linear combination of X1 and X2 with weights that sum to one, that is Π(ω) = ωX1 + (1 − ω)X2 , indicating the proportions of stock 1 and stock 2 that an investor holds. In this example, we require ω ∈ [0, 1], so that negative weights are not allowed. (This rules out short-selling.) (a) Calculate E[Π(ω)] in terms of ω, µ1 and µ2 . Solution: E[Π(ω)] = E[ωX1 + (1 − ω)X2 ] = ωE[X1 ] + (1 − ω)E[X2 ] = ωµ1 + (1 − ω)µ2 (b) If ω ∈ [0, 1] is it possible to have E[Π(ω)] > µ1 and E[Π(ω)] > µ2 ? What about E[Π(ω)] < µ1 and E[Π(ω)] < µ2 ? Explain. Solution: No. If short-selling is disallowed, the portfolio expected return must be between µ1 and µ2 . (c) Express Cov(X1 , X2 ) in terms of ρ and σ1 , σ2 . Solution: Cov(X, Y ) = ρσ1 σ2 (d) What is V ar[Π(ω)]? (Your answer should be in terms of ρ, σ12 and σ22 .) Solution: V ar[Π(ω)] = V ar[ωX1 + (1 − ω)X2 ] = ω 2 V ar(X1 ) + (1 − ω)2 V ar(X2 ) + 2ω(1 − ω)Cov(X1 , X2 ) = ω 2 σ12 + (1 − ω)2 σ22 + 2ω(1 − ω)ρσ1 σ2 (e) Using part (d) show that the value of ω that minimizes V ar[Π(ω)] is ω∗ = σ22 − ρσ1 σ2 σ12 + σ22 − 2ρσ1 σ2 In other words, Π(ω ∗ ) is the minimum variance portfolio. Page 6 Solution: The First Order Condition is: 2ωσ12 − 2(1 − ω)σ22 + (2 − 4ω)ρσ1 σ2 = 0 Dividing both sides by two and rearranging: ωσ12 − (1 − ω)σ22 + (1 − 2ω)ρσ1 σ2 = 0 ωσ12 − σ22 + ωσ22 + ρσ1 σ2 − 2ωρσ1 σ2 = 0 ω(σ12 + σ22 − 2ρσ1 σ2 ) = σ22 − ρσ1 σ2 So we have ω∗ = σ22 − ρσ1 σ2 σ12 + σ22 − 2ρσ1 σ2 (f) If you want a challenge, check the second order condition from part (e). Solution: The second derivative is 2σ12 − 2σ22 − 4ρσ1 σ2 and, since ρ = 1 is the largest possible value for ρ, 2σ12 − 2σ22 − 4ρσ1 σ2 ≥ 2σ12 − 2σ22 − 4σ1 σ2 = 2(σ1 − σ2 )2 ≥ 0 so the second derivative is positive, indicating a minimum. This is a global minimum since the problem is quadratic in ω. Page 7