Scribd
Upload
63 views

ST 522-001: Statistical Theory II Solution To Lab Exercises - 4

This document contains solutions to statistical theory lab exercises. It discusses properties of distributions and sufficient statistics. Specifically, it addresses: 1) When parameter λ equals 0 or 1, the family is not complete. 2) Minimal sufficient statistics for mixtures of normal and gamma distributions involve sample means. 3) The sufficient statistic for mixtures of two uniform distributions is the largest order statistic. 4) Sample means and variances are independent under normal models, and variances are ancillary for means. Order statistics are independent of sufficient statistics for scale families.

Uploaded by

Jonathan Hu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
63 views

ST 522-001: Statistical Theory II Solution To Lab Exercises - 4

This document contains solutions to statistical theory lab exercises. It discusses properties of distributions and sufficient statistics. Specifically, it addresses: 1) When parameter λ equals 0 or 1, the family is not complete. 2) Minimal sufficient statistics for mixtures of normal and gamma distributions involve sample means. 3) The sufficient statistic for mixtures of two uniform distributions is the largest order statistic. 4) Sample means and variances are independent under normal models, and variances are ancillary for means. Order statistics are independent of sufficient statistics for scale families.

Uploaded by

Jonathan Hu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

ST 522-001: Statistical Theory II

Solution to Lab Exercises - 4


Prepared by Chen-Yen Lin
Feb. 09, 2011

6.24; (i) When = 0, random variable X degenerates to 0. It follows immediately that


this family is not complete. (ii) When = 1, to show this family is not complete,
we can let g(0) = e and g(X) = 1, x > 0. Then it can be shown that Eg(X) =
P
h(x)
g(0)e1 + e1
x=1 x! = 1 + 1 = 0.
6.26: (a) If P0 consists two distributions N (1 , 1) and N (0 , 1). By Theorem 6.6.5, a minimal
sufficient statistic for P is
" n
#
2
n1
X
p1
e 2
T (x) =
=
exp
xi (1 0 )
n 2
p0
20
i=1
e

which is equivalent to X
(b) Similarly, consider a case in which P0 consists only two distributions gamma(, 1 )
and gamma(, 0 ). By Theorem 6.6.25, a minimal sufficient statistic for P0 is
" n
#
X 1
1
0
xi

T (x) = exp
1

1
0
i=1
P
which is equivalent to ni=1 Xi .
(c) Consider a case in which P0 consists only two distributions U(0,1 ) and U(0,0 ).
By Theorem 6.6.5, the minimal sufficient statistic for P0 is
T (X) =

1n I(X(n) ,) (1 )
0n I(X(n) ,) (0 )

which is X(n)
is complete sufficient statistic for ,
6.31(a) (i) When 2 is known, it can be shown that X
whereas the distribution of (n 1)S 2 / 2 does not depend on and therefore 2 is
S 2.
ancillary for . By Basu Theorem, X
P
2=
(ii) When P
2 is unknown but fixed, first observe that S 2 = (n 1)1 ni=1 (Xi X)
n
2 , where Zi = Xi is N (0, 2 ). Therefore, S 2 is still an
(n 1)1 i=1 (Zi Z)
S 2 for unknown and fixed 2 . Since 2 is
ancillary statistic. By Basu Theorem, X
arbitrary, this result can be extended to general unknown and 2 . (More discussion
can be found in Example 6.2.27.)
1

6.31(c) (i) Given that X/Y and Y are independent, we have



k
X
EX = E Y
Y
 k
X
k
k
EX = EY E
Y
 k
k
EX
X
= E
k
EY
Y
P
(ii) Since is known, it can be argued that T = ni=1 Xi is complete sufficient for
. Moreover, since Xi comes from a scale family with standard variable Zi whose
X
distribution does not depend on . It follows that Pn (i)Xi is ancillary for . Therefore
k

i=1


E(X(i) |T ) = E

X(i)
T |T
T


=E

X(i)
|T
T


E(T |T ) = E

The last equality holds by using the result from (i).

X(i)
T


T =

E(X(i) )
T
ET

You might also like