Math 362, Problem set 10

Due 4/25/10 (okay to turn in on 4/27 - I will be traveling, though) 1. (7.5.10) Let X1 , . . . , Xn be a random sample from a distribution of pdf f (x; ) = 2 xex . (a) Argue that Y = Xi is a complete sucient statistic for .

(b) Compute E[1/Y ] and nd the function of Y which is the unique MVUE of . Answer: We have f (x; ) = ex+2 log()+log(x) Since this is a regular exponential family, we have that is a complete sucient statistic. We note that Xi (2, 1/) and hence Y (2n, 1/). Thus E[1/Y ] 2n y 2n2 ey dy (2n 1)! 0 (2n 2)! 2n1 2n2 y = y e dy (2n 1)! 0 (2n 2)! = , 2n 1 =

Xi =

K (x)

where we used the fact that the integral in the second line is integrating the pdf of a (2n 1, ) random variable over its entire domain. Thus 1 2n1 E[ 2n is an MVUE of . Y ] = and Y 2. (8.1.2) Let the random variable X have the pdf f (x; ) = (1/)ex/ for 0 < x < , zero elsewhere. Consider the simple hypothesis H0 : = = 2, and the alternative hypothesis H1 : = = 4. Let X1 , X2 denote a random sample of size 2 from this distribution. Show that the best test of H0 against H1 may be carried out by use of the statistic X1 + X2 . Answer:

We have L(0 ) 1 = e(X1 +X2 )/0 +(X1 +X2 )/1 = 2e(X1 +X2 )/4 . L(1 ) 0 Since this is a decreasing function of X1 + X2 the test can be performed by accepting H1 if X1 + X2 c. 3. (8.2.8) Let X1 , . . . , Xn denote a random sample from a normal distribution N (, 16). Find the sample size n and a uniformly most powerful test of H0 : = 25 against H1 : < 25 with power function () so that approximate (25) = .10 and (23) = .9. Answer Note that L(0 ) L(1 ) = = exp 1 1 (Xi 0 )2 (Xi 1 )2 32 32 1 1 2 2 exp (1 0 ) Xi (n0 n1 ) . 16 32 Xi , we accept H1 if
Xi n 4 n

Since this is an increasing function of We have that

Xi < c.

Xi N (n, 16n), and hence

N (0, 1).

We now wish to nd n so that P=25 ( P=23 ( Xi > c) Xi > c) = .1 = .9

This amounts to solving the equations 1.28 1.28 = = c 25n 4 n c 23n 4 n

Thus subtracting the rst from the second, 2 n 2.58 = 4 Or n 26.62. Thus n = 27 suces. 4. (8.2.11) Let X1 , . . . , Xn be a random sample from a distribution with pdf f (x; ) = x1 , 0 < x < 1, zero elsewhere, where > 0. Find a sucient statsitic for and show that a uniformly most powerful test of H0 : = 6 against H1 : < 6 is based on this statistic. Answer: 2

We found such a statistic on the last homework to be L(6) 6 = ( L(1 ) 1 Xi )61 .

Xi . We have

For 1 < 6, this is an increasing function of Xi . That is, this test is a function of Xi and we accept H1 if Xi < c.