Statistics 522 Midterm 2 April 16, 2014 Name: _________________________________ (please print) There are 3 questions. Show your work for full credit. You may assume well known theorems for exponential families that were covered in class. The general form of exponential families is given below. If using such theorems, you need to show that the conditions (if any) hold. Confidence intervals should be in a form where the parameter is between two random quantities. For your information: 1. The Cramer-Rao inequality for the variance of an unbiased estimator of is VarW X 2 2 E log f X . 2. The general form of an exponential family is f x h x c exp k i 1 wi ti x , where h x 0 and c 0 . 3. Definition: A family of pdfs or pmfs f x with real-valued parameter has monotone likelihood ratio (MLR) in the statistic T x if for every 1 0 , whenever f x 1 and f x 0 are distinct, the ratio f x 1 f x 0 is a monotone function of T x for the set of values x for which at least one of the pdfs or pmfs is positive. 1 , X n be a random sample from N , 2 . 1. (30 points total) Let X 1 , 2 (a) (10 points) Show that X , S 2 is a complete sufficient statistic for , 2 . We have f x , n 1 exp 2 n i 1 xi 2 / 2 2 n n x 2 n x n 2 1 i1 i i1 i . exp 2 2 2 2 2 2 1 1 , w1 2 , w2 2 2 , 2 This is an exponential family, with h x 1 , c n T1 x i 1 xi , and T2 x i 1 xi2 . Then since w1 , w2 contains an open set in n n , 2 varies, function of n n 2 as x , i 1 xi2 is complete and sufficient. Since X , S 2 is a one-to-one n i 1 i x , i 1 xi2 , it is complete and sufficient as well. i 1 i n 2 2 We now show that S is UMVUE for (give details). Since S 2 2 2 2 2 n1 , E S . Since n 1 S 2 is unbiased for 2 and a function of the complete sufficient statistic, it is UMVUE. (b) (10 points) Determine the UMVUE for 2 . E X 2 Var X E X 2 2 n 2 . Then 2 n Xi X 2 n 1 2 2 S2 2 2 2 i 1 E X E X S S E X S n n n n X i2 E i 1 S 2 2 . This statistic is unbiased and a function of the complete sufficient statistic, n and is thus the UMVUE for 2 . 2 (c) (10 points) The maximum likelihood estimator of is W X 2 1 n X X . Assume that i n i 1 D n W X 2 N 0, ???? holds. Determine regularity conditions are satisfied so that the result the asymptotic variance, i.e. replace the ????. Completely justify your answer. 2 When the regularity conditions hold, MLEs are asymptotically efficient. Thus ???? is 1/ I , the Fisher information from one observation. 2 I E 2 log f X E 2 2 log f X . 2 2 Now, l , 2 log f x , 2 2 l , 2 2 2 2 x . 1 log 2 2 2 2 2 2 2 1 l , 2 2 2 1 4 x 2 2 2 2 x 2 3 2 . 2 1 2 1 . E 2 log f X 4 3 4 2 2 2 4 Thus ???? 2 . 2. (50 points total) Let X1 , X n be a random sample from the probability density function f x 2x 2 , 0 x . (a) (10 points) Show that X n max X1 , , X n is a consistent estimator sufficient for . x FX x 0 2t x2 2 E X n 0 dt 2 2nx 2 n 2n , so that FX n x dx x2n 2n and f X n x 2nx 2 n 1 2n 2n 2 n 1 2n as n . Also 2n 2n 1 2n 1 3 , 0 x . Then 2nx E X 2 n 2 n 1 2n dx 0 Var X 2 n 2n 2 n 2 2n n 2 2 2n n 1 2n 2 2n 2 so that n 2 2n n 2 0 as n . Thus the bias and 2 n 1 n 1 2n 1 2n 1 2 variance go to zero as n , a sufficient condition for consistency. (b) (5 points) Show that Let Y X X n is a pivotal quantity. y . Then X Y , J , and fY y f X y 2 y , 0 y 1 . FY y 2tdt y 2 and thus FY y y . Differentiating we get fY y 2ny 2n n Y n X n n 0 2 n 1 for 0 y 1 . Thus is then a pivotal quantity. (c) (10 points) Using the pivotal quantity for with lower limit L X X n . X n above, derive a 1 100% confidence interval b 1 for some 0 a b 1 . The confidence interval is then of the form X n X n C x n : . Then b 1 since the lower bound of the interval has been specified b a 1 X n 1 1 . Thus 2ny 2 n 1dx 1 a 2 n 1 . Then a 1/2n and as X n . Therefore Pr a a X n the interval is C X n : X n 1/ 2 n . We set Pr a X n (d) (10 points) Derive a 1 100% confidence interval by pivoting using the cumulative distribution function of X n . n x2n FX n x P X n x P X 1 x 2 n . The interval is then 2n 2n X n X n X 2nn X n X n 2n . Here 1 2 . : 1 2 n 1 2 : : 1/ 2 n 1/ 2 n 1 1 2 1 1 2 4 (e) (10 points) Obtain a likelihood ratio test of H 0 : 0 versus H1 : 0 with level of significance . The likelihood function is not differentiable, but graphically we see that X n is the MLE of . The likelihood ratio statistic is 2n x1 x2 x xn I X n 0 2n 0 n 2 x1 x2 xn x2nn we always reject if X n 0 x n 0 if x n 0 X n 0 or if 0 2n if x n 0 . We reject for small values of x , i.e. 2n 2n 2n 1/2 n c X n 0 c , i.e. X n 0 c . x 2n Then PH X n 0c1/2n PH X n 0 . The cdf of X n is P X n x FX x . 0 Then PH X n 0c 0 1/2 n 0 P X H0 0 n n 2n c1/2 n 0 0 c , and thus we reject if X n 0 or if 0 X n 0 1/2n . (f) (5 points) Invert your test from part (e) to obtain a 1 100% confidence interval for . The non-rejection region is of the form x : 0 1/2n X n 0 . The confidence interval is then of the form : 1/ 2 n X n : X n 3. (20 points total) Let X1 , X n . 1/ 2 n X n be a random sample from a distribution with pdf f x x 1 , 0 x 1 , 0 . (a) (10 points) Use the Neyman-Pearson Lemma to derive the form of the most powerful test for testing H 0 : 0 vs. H1 : 1 , 0 1 . (Your rejection region should depend on the sufficient statistic.) Is your test most powerful for testing H 0 : 0 vs. H1 : 0 ? Why or why not? 5 We calculate the ratio of the pdfs: function of T x x1 x2 0 x1 x2 xn 0 1 x1 x2 xn 1 1 1 0 x1 x2 1 xn 0 1 . This is an increasing xn since 0 1 . We then reject when T x k . Since the form of the test does not change, it is UMP. (b) (5 points) Find explicitly the critical value k for testing H 0 : 0 vs. H1 : 1 when n 1 , 0 0.5 , 1 0.25 , and 0.05 . x 1 k We reject if 0 1 2 x 0.50.25 k , i.e. x 1 x 2 4 0 1 rejection region is then reject if x k / 2 4 0 1 1/ 2 k2 x dx 0.05 0.05 k 0.20 . The 2 4 4 k 0.04 0.0025 . 16 16 (c) (5 points) Under the conditions of the test from part (b), determine the power function the test that you derived. Is your test unbiased? Why or why not? We reject if x 0.0025 . The power function is P x 0.0025 0.0025 x 1dx 0.0025 . Since 0 this is a decreasing function of , it is larger than 0.05 (the power under the null hypothesis) when 0.5 0 . Thus the test is unbiased. 6
© Copyright 2025 Paperzz