This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: m'ieme'mmﬁmj sup WU my )
6500. 11 a m
sgp L(9y1.m.ym) ' 8.3 The LRT statistic is Air) == Let yI= lyi and note that the MLE in the numerator is the minimum of (ﬁrm, 30)
1:: _
(see Exercise 7.12) while the denominator hes y/m an the MLE (see Example 7.2.5). Thus 19 if 34m 5 901
(99)’(1 " 5’0)”,y (i’lmiS (1 — VIMEH' . 1' .. m'y . .
(90) (1 90) < c. To show that this is e"l‘li‘mlﬁm' to Iquims d we re'ect B if ..
a" J 0 (was) Sky/m) _ .
if y > b, we could show My 15 decreasmg m is easier to work with 103 My) (which is equivalent) and we have 103%!) = W139“ + (111 ~ y} 103(1 ' 99)  ¥1°g(%)‘ (:11 * 3r)1°s( Ay :
() y so that My) < (2 occurs for y > b > 12199. It m—Y
T and F 3
5%?) + (m  105:? 1
d log My) 2:: log 60 —— 10g (1  60)  iogG—é) — yy + log( "<5
my
59 ("W )
(In—yum < 1  99, so eadi fraction above is less than i, and the i For y/m > 90. i —y/m =
103 A < 0, which shows that A is decreasing in y and My) <1: if Elogiaieestimnﬂ. Thnsac‘f
andonlyify>b. 8.5 a) The loglikelihood is 7
log L(9'px) = n 1030 + n6 logy — (0+1)log( Hxi), v 5 1(1), where 3(1) = For any vaiue of 6, is an increasing function of v for v 5 1(1). Se both the restricted and unrestricted MLE of v is 5' = :0). To find the MLE of 0, set
3 ' _., 9. _, ', _
wiog L(9,x(1)x) — a + niogxu) iog( H35) — O and solving for 9 yields
 Ii  _ n L
" “’i'o'stnxi/Egy T 7 (62/892) iog L(6,x(1)]x) .1: "11/92 < o, for oil 9. So a is a maximum.
b) Under Ho, the MLE of 9 is £20 = 1 and theMLE of v is still 9 23:”). So the likelihood ratio statistic is  ~
aria,“ {1:92 A E
(n/T)“x?1)/T/ ( [hon/T *1
(6/8T)log M1) = (ll/T)  1. So M1) is increasing if T s n and decreasing if '1'; n. Thus, T s c is equivalent to T _<_ :1 or T 3 c2, for appropriateiy chosen constants c1 and c2. 8.6 a) 11) “sup L(9li) su‘p ﬁie'xi/a ﬁlc'Vi/e
9 i=1‘9 i=1” a
o
Kuhn L9 = n “a ,_
9p (lwi ﬂup.111 :57 111317;: _._‘ e
0,}: 1:19 i=1 n 'm
.. 9
sup 1 8 (1:1 121 1/ 9 9111411
= W‘
 E Jig/9  B H.“
.1. i=1 1 i=1 1
ii: 25° Diﬁ'erentiation show that in the numerator 30 = (21:1 + mi)/(n+m), while in the denominator 3 2 5E and it = i. i '  e .‘
Therefore. Many) =_ + l = (n+m)n+m (mnwygm nﬁEm + ﬂyinm‘
And the LRT is to reject H9 if Mr, y) S c. n m
A _ (n+m)n+m Bari __ (n+m)n+m
“ nﬁﬁﬁ 2f, + E E, +Eyi " 115551
Timion A is a function of T. A is a unimadal function orrr which is maximized when T = min. Rejection for A S c is equivalent to rejection {or T 5 a or T 2 b, where a and b are constants that satisfy an(Ia)m a bun—b)”. Tn(1 —T)'n. When He is true, ZXi ~ gamma(n,9} and EYi ~ gamma.(m,9) and they are independent.
So by an extension of Exercise 4.192), T ~ beta.(n,m). L?
w a.) The likelihood function is
Maﬁa!) = #“(l'llq)”"19“(llyi)9“l Lannimaﬁwnxinygal, and maximizing as above yields the restricted MLE, a 3 w n + m M
0 Slogan?» Slosyi'
The LET statistic is ,
' am“ a _‘ a 5
. A(IJ)=E‘.’r§~a,a(llx;) 9 “(Evil 0  n ‘ _
1:) Substituting in the formulas for b.‘ :1, and 30 yields ([1:990” “(nyifo’ 9
A( ) _,,. £314.11 _ 131+“ m(m+tl)n T an
I an! if‘nrem—Eﬁgﬁ"? (l 
So rejecting if MI, 3!) 5 c is equivalent to rejecting if aloud This is a unimodal function of T.
T 5 c1 or T 2 c2, where c1 and :2 are appropriately chosen constants.
c) A simple transformation yields r—log Xi ~ exponentialﬂ/p) and logYi ~ exponential(1/ 0). Therefore, T 2 W/(W+V) where W and V are independent, W ~ gmMn,1/p), and A V ~ gmmﬂmJ/B). Under 130, the scale parameters of W and V are equal. Then, a simple generalization of Exercise 4.1§b yields T we bote{n,m). The constants c; and c2 are
(ier by the two equations ‘ .
. y Pﬁswpazem and (lcl)“‘c‘£=(Ic2)‘“c‘2‘ a“? r A
a —P lineal w1 P m‘6"l<
z 1 “PAﬁiEg i—oo 5%) , 7 7_
“ ca/'~l'ﬁ+30—3 iﬁg cor/«I’ﬁiBgE
«1—P9( a/ﬁ 01113
9 9 0 —0
._ m .. .0 ' 9 ~
#1 P( c+amszgc+alﬁ)_ (z mm)
90 809 _
(Q is the standard normaledf) b) The size is .05 = 5096) = 1 + @(—e) w<1>(c) z: c = 1.96. The power (1 — type 11 error) is .75 g 13(90—1—0) = 1 + @(c~i§)¢(c—‘Tﬁ) 2 1 + owes—r) —‘1’(1.96\[ﬁ).
“NIH—I'd . #0 From Table 1, Q(.675) to .25 => 136—411" m—.675'=> n = 6.943 $3 7. ...
View
Full Document
 Spring '07
 SENTURK,DAMLA
 Statistics

Click to edit the document details