STAT 265 Pre-Final ID: Name: 1. Let 𝑌1 , 𝑌2 , … , 𝑌𝑛 be a random sample of a binomial distribution with n trials and success probability p. 1 a. Show 𝑌̅ = 𝑛 ∑𝑛𝑖=1 𝑌𝑖 is a biased estimator of p b. Show 𝑌𝑖 𝑛 is a consistent estimator of p Hint: 𝐸(𝑌1 ) = 𝐸(𝑌2) = ⋯ = 𝐸(𝑌𝑛 ) = 𝑛𝑝, 𝑉(𝑌1 ) = 𝑉(𝑌2 ) = ⋯ = 𝑉(𝑌𝑛 ) = 𝑛𝑝(1 − 𝑝) 𝑉(𝑎𝑌) = 𝑎2 𝑉(𝑌) For part (b), Use the theorem: If 𝜃̂𝑛 is an unbiased estimator for θ and lim 𝑉(𝜃̂𝑛 ) = 0, 𝜃̂𝑛 is a consistent estimator of 𝑛→∞ θ. This means you need to show two things - 𝑌𝑖 𝑛 is an unbiased estimator 𝑌 lim 𝑉 ( 𝑖 ) = 0 𝑛 𝑛→∞ 2. Let 𝜃̂1 , 𝜃̂2 be unbiased estimator for a parameter θ where E(𝜃̂1 ) = 𝐸(𝜃̂2 ) = 𝜃, 𝑎𝑛𝑑 𝑉(𝜃̂1 ) = 𝜎12 , 𝑉(𝜃̂2) = 𝜎22 Consider another estimator 𝜃̂3 = 𝑎𝜃̂1 + (1 − 𝑎)𝜃̂2 , 𝑎 ∈ ℝ. a. Show 𝜃̂3 is an unbiased estimator for θ b. Suppose 𝜃̂1 , 𝜃̂2 are independent. Find the variance of 𝜃̂3 , 𝑉(𝜃̂3). What the value of a minimize 𝑉(𝜃̂3 )? Show your all work. Hint: For par b, 𝜃̂1 , 𝜃̂2 are independent ⇒ 𝐸(𝜃̂1 𝜃̂2 ) − E(𝜃̂1 )𝐸(𝜃̂2) = 𝐸(𝜃̂1 𝜃̂2 ) − 𝜃 2 = 0 f ’’(x)>0 for any x, f(x) is convex. If f(x) is convex, f(x) will be minimized when f ’(x)=0 3. Suppose Y1 , Y2 , Y3 are independent random variables which follow the uniform distribution, U(0, θ). 𝐸(𝑌1 ) = 𝐸(𝑌2 ) = 𝐸(𝑌3 ) = 𝜃 , 2 𝑉(𝑌1 ) = 𝑉(𝑌2 ) = 𝑉(𝑌3 ) = 𝜃2 12 Consider the estimators of θ: 𝜃̂1 = 𝑌1 , 𝜃̂2 = 𝑌1 + 2𝑌2 , 3 𝜃̂3 = 𝑌̅ = 𝑌1 + 𝑌2 + 𝑌3 3 a. Find the variances of 𝜃̂1 , 𝜃̂2 , 𝜃̂3 b. Find the efficiency of 𝜃̂2 relative to 𝜃̂3 Hint: If X, Y are independent and a, b are constant, V(aX+bY)=a2 V(X)+b2 V(Y) Efficiency of 𝜃̂2 relative to 𝜃̂3 is given by 𝑉(𝜃̂3 ) 𝑉(𝜃̂2 ) 4. Let 𝑌1 , 𝑌2 , … , 𝑌𝑛 denote a random sample from the probability density function given by (𝜃 + 3)𝑦 𝜃 , 0 < 𝑦 < 1; 𝜃 > −3 𝑓(𝑦|𝜃) = { 0, 𝑒𝑙𝑠𝑒𝑤ℎ𝑒𝑟𝑒 a. Find an estimator for θ by the method of moments. b. Find an estimator for θ by the method of maximum-likelihood. Show your all work. Hints: For the part b, L(θ) is maximized when ln[L(θ)] is maximized with the derivative dln[L(θ)]/dθ=0 5. Extra Question (you would gain extra marks if you get the correct answers) Population is normally distributed with unknown μandσ=1. Let 𝑌̅ represent a sample of size n. The sampling distribution with n=9 has the probability that the sample mean will be within 0.3 of μ, P(|𝑌̅ − 𝜇| ≤ 0.3) = 0.6318. Show your work. a. If n=16, what isP(|𝑌̅ − 𝜇| ≤ 0.3)? b. What is P(|𝑌̅ − 𝜇| ≤ 0.3), if n=25, 36? Hint: For sampling distribution, z = 𝑌̅−𝜇 𝜎/√𝑛 𝑌̅−𝜇 = √𝑛 ( 𝜎 ) −0.3√𝑛 𝑌̅ − 𝜇 0.3√𝑛 −0.3√𝑛 0.3√𝑛 P(|𝑌̅ − μ| ≤ 0.3) = P ( ≤ √𝑛 ( )≤ ) = P( ≤𝑍≤ ) 𝜎 𝜎 𝜎 𝜎 𝜎 P(−0.9 ≤ Z ≤ 0.9) = 0.6318 P(−1.2 ≤ Z ≤ 1.2) = 0.7699 P(−1.5 ≤ Z ≤ 1.5) = 0.8664 P(−1.8 ≤ Z ≤ 1.8) = 0.9281 1. Solutions 1 1 1 a. 𝐸(𝑌̅) = 𝐸 (𝑛 ∑ 𝑌𝑖 ) = 𝑛 [𝐸(𝑌1 ) + ⋯ + 𝐸(𝑌𝑛 )] = 𝑛 ∙ 𝑛(𝑛𝑝) = 𝑛𝑝, so a biased estimator 𝑌 1 1 b. 𝐸 ( 𝑛𝑖 ) = 𝑛 𝐸(𝑌𝑖 ) = 𝑛 ∙ 𝑛𝑝 = 𝑝, so an unbiased estimator 𝑌𝑖 1 1 𝑝(1 − 𝑝) 𝑉 ( ) = 2 𝑉(𝑌𝑖 ) = 2 ∙ 𝑛𝑝(1 − 𝑝) = → 0 as 𝑛 → ∞ 𝑛 𝑛 𝑛 𝑛 2. Solutions: a. 𝐸(𝜃̂3) = 𝐸(𝑎𝜃̂1 + (1 − 𝑎)𝜃̂2 ) = 𝑎𝐸(𝜃̂1 ) + (1 − 𝑎)𝐸(𝜃̂2 ) = 𝑎𝜃 + (1 − 𝑎)𝜃 = 𝜃 2 b. 𝑉(𝜃̂3 ) = 𝐸(𝜃̂32 ) − [𝐸(𝜃̂3 )] = 𝐸(𝑎2 𝜃̂12 + 2𝑎(1 − 𝑎)𝜃̂1 𝜃̂2 + (1 − 𝑎)2 𝜃̂22 ) − 𝜃 2 = 𝑎2 𝐸(𝜃̂12 ) + 2𝑎(1 − 𝑎)𝐸(𝜃̂1 𝜃̂2 ) + (1 − 𝑎)2 𝐸(𝜃̂22) − 𝜃 2 +2𝑎(1 − 𝑎)𝜃 2 − 2𝑎(1 − 𝑎)𝜃 2 2 2 = 𝑎2 (𝑉(𝜃̂1 ) + (𝐸(𝜃̂1)) ) + (1 − 𝑎)2 (𝑉(𝜃̂2 ) + (𝐸(𝜃̂2)) ) ⏟ ⏟ ̂ 2) 𝐸(𝜃 1 ̂2) 𝐸(𝜃 2 2 2 +2𝑎(1 − 𝑎) [𝐸(𝜃 ⏟ ̂1 𝜃̂2 ) − 𝜃 ] − 𝜃 (1 − 2𝑎(1 − 𝑎)) 0 = 𝑎2 (𝜎12 + 𝜃 2 ) + (1 − 𝑎)2 (𝜎22 + 𝜃 2 ) − (2𝑎2 − 2𝑎 + 1)𝜃 2 = 𝑎2 [(𝜎12 + 𝜃 2 ) + (𝜎22 + 𝜃 2 ) − 2𝜃 2 ] + 𝑎[−2(𝜎22 + 𝜃 2 ) + 2𝜃 2 ] + [(𝜎22 + 𝜃 2 ) − 𝜃 2] = 𝑎2 (𝜎12 + 𝜎22 ) − (2𝜎22 )𝑎 + 𝜎22 Consider 𝑓(𝑎) = 𝑎2 (𝜎12 + 𝜎22 ) − (2𝜎22 )𝑎 + 𝜎22 . Since (𝜎12 + 𝜎22 ) > 0, this is convex. 𝑓(𝑎) will be minimized when 𝑓′(𝑎) = 0. 𝑓 ′ (𝑎) = 2(𝜎12 + 𝜎22 )𝑎 − 2𝜎22 = 0 𝑎= 𝜎22 2 𝜎1 +𝜎22 3. Solutions 2 𝜃 a. 𝑉(𝜃̂1 ) = 𝑉(𝑌1 ) = 12 𝑌1 + 2𝑌2 1 2 2 2 5 𝜃2 𝑉(𝜃̂2 ) = 𝑉 ( ) = ( ) 𝑉(𝑌1 ) + ( ) 𝑉(𝑌2 ) = ( ) 3 3 3 9 12 𝑌1 + 𝑌2 + 𝑌3 1 2 1 2 1 2 3 𝜃2 𝑉(𝜃̂3 ) = 𝑉 ( ) = ( ) 𝑉(𝑌1 ) + ( ) 𝑉(𝑌2 ) + ( ) 𝑉(𝑌3 ) = ( ) 3 3 3 3 9 12 b. ̂3 ) 𝑉(𝜃 ̂2 ) 𝑉(𝜃 = 3 𝜃2 ( ) 9 12 5 𝜃2 ( 9 12 ) = 3 5 4. a. 1 𝐸(𝑌) = 𝐸(𝐸(𝑌|𝜃)) = ∫ 𝑦𝑓(𝑦|𝜃)𝑑𝑦 = 0 1 ∫0 (𝜃 + 3)𝑦 𝜃+1 𝑑𝑦 𝜃+3 𝜃+3 = (𝜃+2) 𝑦 (𝜃+2) |10 = (𝜃+2) = 𝜇 Solve by θ, θ = 2𝜇−3 1−𝜇 The method-of-moments estimator is 2𝑌−3 1−𝑌 b. 𝐿(𝜃) = 𝑓(𝑦1 |𝜃) × ⋯ × 𝑓(𝑦𝑛 |𝜃) = (θ + 3)𝑦1𝜃 × ⋯ × (𝜃 + 3)𝑦𝑛𝜃 = (𝜃 + 3)𝑛 (𝑦1 ⋯ 𝑦𝑛 )𝜃 𝑛 ln[𝐿(𝜃)] = 𝑛 ln(𝜃 + 3) + 𝜃 ∑ ln 𝑦𝑖 𝑖=1 𝑛 𝑑 ln[𝐿(𝜃)] 𝑛 = + ∑ ln 𝑦𝑖 = 0 𝑑𝜃 𝜃+3 𝑖=1 Solve by θ, 𝜃̂ = −𝑛 ∑𝑛𝑖=1 ln 𝑦𝑖 −3 The Maximum-likelihood estimator is −𝑛 ∑𝑛 𝑖=1 ln 𝑌𝑖 −3 5. Solutions: a. P(| – μ| ≤ .3) = P(–1.2 ≤ Z ≤ 1.2) = .7699. b. P(| – μ| ≤ .3) = P(–.3 ≤ Z ≤ .3 ) = 1 – 2P(Z > .3 ). For n = 25, 36, the probabilities are (respectively) 0.8664, 0.9281.