UNIVERSITY OF LIMPOPO

TURFLOOP CAMPUS

FACULTY OF SCIENCE AND AGRICULTURE

SCHOOL OF MATHEMATICAL AND COMPUTER SCIENCES

DEPARTMENT OF STATISTICS AND OPERATIONS RESEARCH

DEGREE AND DIPLOMA EXAMINATION OCT/NOV: 2016

MODULE: SSTA082 PAPER: P

(STATISTICAL INFERENCE)

TIME: 3 HOURS MARKS: 100

INTERNAL EXAMINER: PROF A TESSERA

EXTERNAL EXAMINER: DR J HUGO

(NMMU)

THIS PAPER CONSISTS OF SIX PAGES INCLUDING THE COVER PAGE.

INSTRUCTIONS:

1. Answer all the questions.

2. Show all your work clearly.
3. Start each question on a new page.
 SSTA082 OCT/NOV EXAMINATION 2016

QUESTION 1 [30]

a) Explain what is meant by saying S  s  X  is a complete sufficient statistic. (4)

b) Let X   X 1 , X 2 , ..., X n  be a random sample from f  .;  and S a sufficient statistic.

If T is any unbiased estimator of    and T   E T S  , then show the following.

i) T  is an unbiased estimator of    . (2)

ii) Var T    Var T  for all  . (4)

iii) If S is a complete sufficient statistic, then T   E T S  is the only unbiased

estimator of    that is a function of S and, therefore, it is a uniformly minimum
variance unbiased estimator (UMVUE) of    . (4)

c) Consider a random sample of size n from the geometric density:

  1   
 x 1
x  1,2,3,...
f  x;   

0 otherwise.

Find:

i) An unbiased estimator of      1    based on X 1 . (4)

ii) A complete sufficient statistic for  . (3)

n
iii) The distribution of S   X i . (3)
i 1

iv) A UMVUE of      1    by conditioning on a complete sufficient statistic.

(6)

QUESTION 2 [20]

a) Suppose that X   X 1 , X 2 , ..., X n  is a random sample from a population with

probability density function f X  .;  where  is unknown.

i) Explain what is meant by a pivotal quantity. (2)

ii) Show how a pivotal quantity may be used to construct a confidence interval for
. (2)

Page 2
 SSTA082 OCT/NOV EXAMINATION 2016

b) Suppose X   X 1 , X 2 , ..., X n  is a random sample from a distribution with density

 1  
 1 x 0 x1
fX  x  

0 elsewhere
where  >0 is unknown.
2
i) Show that Y   ln X has an exponential distribution. What is the mean of the

distribution? (8)
n
2
ii) Show that Q  

 ln X
1
is a pivotal quantity for  . (3)

iii) Construct a 100  1    % confidence interval for  . (5)

QUESTION 3 [25]

a) Define the following terms.

i) Most powerful (MP) test (2)
ii) Uniformly most powerful (UMP) test (2)

b) Let X   X 1 , X 2 , ..., X n  be a random sample from a density that belongs to the one
parameter exponential family; i.e. f  x;   a   b  x  exp c   d  x  where c  . is an
increasing function. Show that the critical region of a UMP test for testing H0 :   0
 n

against H0 :   0 is given by C   x : t  x    d  xi   k  . (6)
 i 1 

c) Let X   X 1 , X 2 , ..., X n  be a random sample of size n from POISSON   .

i) Show that the MP test of H0 :   1 versus H1 :   2 leads to a critical region of

 n

the form  x :  xi  c  . By using the Normal approximation to the Poisson
 i 1 
distribution, determine an approximate critical region for n  36 and   0.05 .
n
(You may use: X
i 1
i ~ N  n , n  ) (10)

ii) Determine an approximate critical region for a UMP test of H0 :   1 versus

H1 :   1 for n  36 and   0.05 . (5)

Page 3
 SSTA082 OCT/NOV EXAMINATION 2016

QUESTION 4 [25]

a) Let X   X 1 , X 2 , ..., X n  be a random sample from the density f  x;  and suppose
we want to test H0 : 0 versus H1 :   0 . Describe the generalised likelihood
ratio test of H 0 against H 1 . State the asymptotic distribution of 2ln  under the null
hypothesis, where  is the test statistic. (5)

b) Consider two independent random samples from two populations. Let

X   X 1 , X 2 , ..., X m  be a random sample of size m from

 2 xe x x 0
fX  x  
0 elsewhere

Let Y  Y1 , Y2 , ..., Yn  be a random sample of size n from

  2 ye  y y 0
fY  y   
0 elsewhere

Suppose we want to test H0 :    against H1 :    . Show that the generalised

likelihood ratio test has critical region of the form

  n
   n

 m   Yi   n   X i 
4mlog   1  in1    4nlog   1  i n1    k .
m  n  X i   m  n  Yi  
 

   

 
 i 1  i 1

n n
Perform the test at the 5% level for m  100 , n  200 , x
i 1
i  800 and y
i 1
i  1500 .

(20)

Page 4
 SSTA082 OCT/NOV EXAMINATION 2016

TABLE OF IMPORTANT DISTRIBUTION FUNCTIONS

NAME TYPE DENSITY FUNCTION MGF MEAN VARIANCE

1
UNIFORM DISCRETE

fX  x   N
if x  1,2,...,N et ( 1  etN ) N 1 N 2
 1
 N( 1  et ) 2 12
o otherwise.
 n  x n  x
 x  0,1,2,...,n
f X  x    x 
pq
BINOMIAL DISCRETE  q  pe  t n np npq
0
 otherwise.
  M  N  M 
  
  x  n  x  for x  0,1,2,...,n M  M  M  N  n 
HYPERGEOMETRIC DISCRETE fX  x   N n n   1   
   N  N  N  N  1 
 n 
0 otherwise.
 x  1  r x  r
x  r,r  ,r  2,...
r
  pet 
f X  x    r  1 
pq r rq
NEGATIVE BINOMIAL DISCRETE  t  p2
0  1  qe  p
 otherwise.
   x
e x  0,1,2,...
POISSON DISCRETE fX  x   x! e
 
 et  1
 
0
 otherwise.

Page 5
 SSTA082 OCT/NOV EXAMINATION 2016

TABLE OF IMPORTANT DISTRIBUTION FUNCTIONS

NAME TYPE DENSITY FUNCTION MGF MEAN VARIANCE

 1
 if a  x  b ebt  e at 1 1
fX  x  b  a  a  b b  a 
2
UNIFORM CONTINUOUS
 b  a  t 2 12
0 otherwise.
 1 r 1

x

 x e 
x 0
f X  x    r  r   1  t 
r
GAMMA CONTINUOUS r r 2
0
 otherwise.
 1  x  
 e x 0
fX  x    1  t 
1
EXPONENTIAL CONTINUOUS  2

0 otherwise.
1 x 
2
 
fX  x 
1 

1
t   2 t 2
NORMAL CONTINUOUS e 2     x   e 2 2
 2
     x  1e    x  0


 x
 1   2 2 1 
fX  x     1   2   1      1  
 
WEIBULL CONTINUOUS
       
0 otherwise.
   1
  x
 1 x 0   2
PARETO CONTINUOUS f X  x      
 1   2   1
2
0
 otherwise.

Page 6