Autumn 08 Stat 620 Solutions for Homework 5 Part II 4.17 (a) P (Y = i + 1) = R i+1 i e−x dx = (−e−x )|i+1 = e−i (1 − e−1 ), which is geometric with p = 1 − e−1 . i (b) Since Y = i + 1 if and only if i ≤ X < i + 1, Y ≥ 5 ⇔ X ≥ 4. P (X − 4 ≤ x|Y ≥ 5) = P (X − 4 ≤ x|X ≥ 4) By the “memoryless” property (on page 101 of Casella & Berger) of the exponential distribution: P (X > s|X > t) = P (X > s − t), P (X − 4 ≤ x|X ≥ 4) = P (X ≤ 4 + x|X ≥ 4) = 1 − P (X > 4 + x|X ≥ 4) = 1 − P (X > 4 + x − 4) = 1 − P (X > x) = P (X ≤ x) = 1 − e−x , x ∈ {0, 1, 2, . . .} 4.20 (a) This transformation is not one-to-one because you cannot determine the sign of X2 from Y1 0 = {−∞ < x < ∞, x = 0}, S 1 = and Y2 . So partition the support of (X1 , X2 ) into SX 1 2 X 2 = {−∞ < x < ∞, x < 0}. The support of (Y , Y ) is {−∞ < x1 < ∞, x2 > 0} and SX 1 2 1 2 0 } = 0. The inverse SY = {0 < y1 < ∞, −1 < y2 < 1}. It is easy to see that P {X ∈ SX p 1 is x = y √y and x = transformation from SY to SX y1 − y1 y22 with Jacobian 1 2 1 2 J1 (y1 , y2 ) = 1 √y2 2 y1 √ 2 1−y2 1 √ 2 y1 √ y2 √ y1 √ y1 1−y22 1 = p 2 1 − y22 p 2 is x = y √y and x = − y − y y 2 with J = −J . The inverse transformation from SY to SX 1 2 1 1 2 2 1 1 2 Then 1 −y1 /(2σ2 ) 1 1 −y1 /(2σ2 ) −1 1 −y1 /(2σ2 ) 1 p p p fY1 ,Y2 (y1 , y2 ) = e + e e = 2 2 2 2 2 2πσ 2πσ 2πσ 2 1 − y2 2 1 − y2 1 − y22 (b) We can see in the above expression that the joint pdf factors into a function of y1 and a function of y2 . So Y1 and Y2 are independent. Y1 is the square of the distance from (X1 , X2 ) to the origin. Y2 is the cosine of the angle between the positive x1 -axis and the line from (X1 , X2 ) to the origin. So 1 Autumn 08 Stat 620 independence says that the distance from the origin is independent of the orientation (as measured by the angle). 4.23 (a) X ∼ beta(α, β), Y ∼ beta(α + β, γ). X and Y are independent. The inverse transformation is y = v, x = u/v with Jacobian J(u, v) = fU,V (u, v) = ∂x ∂u ∂x ∂v ∂y ∂u ∂y ∂v 1 v = 0 −u 2 1 v = 1 v Γ(α + β) Γ(α + β + γ) u α−1 u 1 ( ) (1 − )β−1 v α+β−1 (1 − v)γ−1 , 0 < u < v < 1 Γ(α)Γ(β) Γ(α + β)Γ(γ) v v v So fU (u) = = = = Z Γ(α + β + γ) α−1 1 β−1 v − u β−1 u ) dv v (1 − v)γ−1 ( Γ(α)Γ(β)Γ(γ) v u Z 1 dv Γ(α + β + γ) α−1 v−u β+γ−1 β−1 γ−1 u (1 − u) , dy = y (1 − y) dy y = Γ(α)Γ(β)Γ(γ) 1−u 1−u 0 Γ(β)Γ(γ) Γ(α + β + γ) α−1 u (1 − u)β+γ−1 Γ(α)Γ(β)Γ(γ) Γ(β + γ) Γ(α + β + γ) α−1 u (1 − u)β+γ−1 , 0 < u < 1. Γ(α)Γ(β + γ) Thus, U ∼ beta(α, β + γ). 4.24 Let z1 = x + y, z2 = x x+y , then x = z1 z2 , y = z1 (1 − z2 ) and |J(z1 , z2 )| = ∂x ∂z1 ∂x ∂z2 ∂y ∂z1 ∂y ∂z2 z2 z 1 = = z1 1 − z −z 2 1 The support SXY = {x > 0, y > 0} is mapped onto the set SZ = {z1 > 0, 0 < z2 < 1}. So 1 1 (z1 z2 )r−1 e−z1 z2 · (z1 − z1 z2 )s−1 e−z1 +z1 z2 · z1 Γ(r) Γ(s) Γ(r + s) r−1 1 = z r+s−1 e−z1 · z (1 − z2 )s−1 , Γ(r + s) 1 Γ(r)Γ(s) 2 fZ1 ,Z2 (z1 , z2 ) = 2 0 < z1 < +∞, 0 < z2 < 1. Autumn 08 Stat 620 Z1 and Z2 are independent because fZ1 ,Z2 (z1 , z2 ) can be factored into two terms, one depending only on z1 and the other depending only on z2 . Inspection of these kernals shows Z1 ∼ gamma(r + s, 1), Z2 ∼ beta(r, s). 4.26 (a) The only way to attack this problem is to find the joint cdf of (Z, W ). Now W = 0 or 1. For 0 ≤ w < 1, P (Z ≤ z, W ≤ w) = P (Z ≤ z, W = 0) = P (min(X, Y ) ≤ z, Y ≤ X) = P (Y ≤ z, Y ≤ X) Z zZ ∞ 1 −x/λ 1 −y/µ = e e dxdy λ µ 0 y λ 1 1 = 1 − exp{−( + )z} λ+µ µ λ and P (Z ≤ z, W = 1) = P (min(X, Y ) ≤ z, X ≤ Y ) µ = P (X ≤ z, X ≤ Y ) = λ+µ 1 1 1 − exp{−( + )z} µ λ can be used to give the joint cdf. (b) Z ∞Z ∞ P (W = 0) = P (Y ≤ X) = 0 y λ 1 −x/λ 1 −y/µ e e dxdy = λ µ µ+λ P (W = 1) = 1 − P (W = 0) = µ µ+λ P (Z ≤ z) = P (Z ≤ z, W = 0) + P (Z ≤ z, W = 1) = 1 − exp{−( 1 1 + )z} µ λ Therefore, P (Z ≤ z, W = i) = P (Z ≤ z)P (W = i), for i = 0, 1, z > 0 which implies that Z and W are independent. 4.36 (a) EY = n X i=1 EXi = n X E(E(Xi |Pi )) = i=1 n X i=1 3 E(Pi ) = n X i=1 α nα = α+β α+β Autumn 08 Stat 620 (b) Since the trials are independent, VY = n X VXi = i=1 = n X n X [V(E(Xi |Pi )) + E(V(Xi |Pi ))] = i=1 = [V(Pi ) + E(Pi (1 − Pi ))] i=1 [V(Pi ) − E((Pi )2 ) + E(Pi )] = i=1 n X n X n X [−(E(Pi ))2 + E(Pi )] i=1 [−( i=1 α 2 α ) + ]= α+β α+β n X i=1 αβ nαβ = (α + β)2 (α + β)2 α (c) We show that Xi ∼ Bernoulli α+β . Then, because Xi0 s are independent, Y = Pn i=1 Xi ∼ α ). Let fPi (p) denote the beta(α, β) density for the Pi s and conditioning on Pi gives Binomial(n, α+β 1 Z 1 Γ(α + β) α−1 P (Xi = x|Pi = p)fPi (p) dp = px (1 − p)1−x p (1 − p)β−1 Γ(α + β) 0 0 Z Γ(α + β) 1 α+x−1 = p (1 − p)β−x+1−1 dp Γ(α)Γ(β) 0 β Γ(α + β) Γ(α + x)Γ(β − x + 1) α+β x = 0 = = Γ(α)Γ(β) Γ(α + β + 1) α x=1 α+β Z P (Xi = x) = 4.47 Because the sample space can be decomposed into the subsets [XY > 0] and [XY < 0] (and a set of (X, Y ) probability zero), we have for z < 0, by definition of Z, P (Z ≤ z) = P (X ≤ z and XY > 0) + P (−X ≤ z and XY < 0) = P (X ≤ z and Y < 0) + P (X ≥ −z and Y < 0) (because z < 0) = P (X ≤ z)P (Y < 0) + P (X ≥ −z)P (Y < 0) (because X and Y are independent) = P (X ≤ z)P (Y < 0) + P (X ≤ z)P (Y < 0) (symmetry of the distributions of X and Y ) = P (X ≤ z) By a similar argument, for z > 0, we get P (Z > z) = P (X > z), and hence, Z ∼ X ∼ N (0, 1). (b) By definition of Z, Z > 0 ⇔ either (i) X < 0 and Y > 0 or (ii) X > 0 and Y > 0. So Z and Y always have the same sign, hence they cannot be bivariate normal. 4 Autumn 08 Stat 620 4.58 These parts are exercises in the use of conditioning. (a) Cov(X, Y ) = E(XY ) − E(X)E(Y ) = E[E(XY |X)] − E(X)E[E(Y |X)] = E[XE(Y |X)] − E(X)E[E(Y |X)] = Cov(X, E(Y |X)) (b) Cov(X, Y − E(Y |X)) = Cov(X, Y ) − Cov(X, E(Y |X)) = 0 by (a) So, X and Y − E(Y |X) are uncorrelated. (c) V ar(Y − E {Y |X}) = V ar(E {Y − E {Y |X} |X}) + E {V ar [Y − E {Y |X} |X]} = V ar (E {Y |X} − E {Y |X}) + E {V ar (Y − E {Y |X} |X)} = V ar(0) + E {V ar(Y |X)} = E {V ar(Y |X)} 5 E {Y |X} is constant wrt the [Y |X] distribution