Scribd
Upload
146 views3 pages

No Data Decision Problem Question 1

This document discusses a no-data decision problem with 6 actions and 2 states of nature. It provides the loss function and asks to (a) find inadmissible actions, (b) determine the minimax pure action, (c) find the minimax mixed action, (d) determine the Bayes action for a given prior distribution, (e) identify the least favorable distribution, and (f) distributions where the Bayes action is not unique. The solution identifies a2 and a3 as inadmissible, a6 as the minimax pure action, a mixed action of a5 and a6 as minimax, and the Bayes action depends on the prior probability w. It also provides the least favorable distribution and two prior

Uploaded by

Iam Leong
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
146 views3 pages

No Data Decision Problem Question 1

This document discusses a no-data decision problem with 6 actions and 2 states of nature. It provides the loss function and asks to (a) find inadmissible actions, (b) determine the minimax pure action, (c) find the minimax mixed action, (d) determine the Bayes action for a given prior distribution, (e) identify the least favorable distribution, and (f) distributions where the Bayes action is not unique. The solution identifies a2 and a3 as inadmissible, a6 as the minimax pure action, a mixed action of a5 and a6 as minimax, and the Bayes action depends on the prior probability w. It also provides the least favorable distribution and two prior

Uploaded by

Iam Leong
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/ 3

Leong YK & Wong WY Introduction to Statistical Decisions 1

Chapter 2
No Data Decision Problems
Examples
2.1 Consider a no-data problem with six available actions, two states of
nature, and the loss function given in the following table

a1 a2 a3 a4 a5 a6
1 2 5 4 6 1 3
2 4 3 3 1 5 2

(a) List all the inadmissible pure actions


(b) Find the minimax pure action.

(c) Determine the minimax mixed action.

(d) Suppose that the decision maker prior belief of the states of natureis
given as follows;
 : P(  1)  w , P(  2 )  1  w , 0  w  1.
Determine the Bayes action against  .

(e) What is the least favorable distribution of the decision maker?


(f) Find all the prior distributions so that the Bayes action is not unique.
Leong YK & Wong WY Introduction to Statistical Decisions 2

Solution

(a) Inadmissible pure actions : a2 and a3 .


(b) Mimimax pure action : a6 .
(c) It is easily seen that the minimax mixed action is a mixture of
actions a5 and a6 , say
~
p  p  a5  (1  p)  a6 
where p is determined by equation L(1, ~p )  L( 2 , ~p ) , i.e.,
p  3(1  p)  5 p  2(1  p) or p  1 / 5
1 4
Answer : ~p   a5    a6  .
5 5
(d) The Bayes action against  must be one of the admissible
actions
L( , a1 )  2 w  4(1  w)
L( , a4 )  6w  (1  w)
L( , a5 )  w  5(1  w)
L( , a6 )  3w  2(1  w)
Leong YK & Wong WY Introduction to Statistical Decisions 3

The Bayes action against the prior distribution  is given by


a 4 , 0  w  1/ 4

a  a6 , 1/ 4  w  3 / 5.
a 3/ 5  w  1
 5 ,
(e) The least favorable distribution is
 L : P(  1 )  3 / 5 , P(   2 )  2 / 5
(f) The prior distributions for which the Bayes action is NOT
unique are
1 : P(  1 )  1 / 4 , P(   2 )  3 / 4
 2 : P(  1 )  3 / 5 , P(   2 )  2 / 5

You might also like