LOYOLA COLLEGE (AUTONOMOUS), CHENNAI – 600 034
|
M.Sc. DEGREE EXAMINATION – STATISTICS
SECOND SEMESTER – APRIL 2008
ST 2810 – SAMPLING THEORY
Date : 24/04/2008 Dept. No. Max. : 100 Marks
Time : 1:00 – 4:00
SECTION – A
Answer ALL questions. Each carries TWO marks. (10 x 2 = 20 marks)
- Define Population. I i (s) What are the assumptions made about population size?
- Distinguish between parameter and statistic. Give an example for both.
- Find the following:
(i) E p [ I i (s) ] , i = 1, 2, …, N,
(ii) E p [ I i (s) E p [ I i (s) I j (s)] ; i , j = 1, 2, …, N ; i ≠ j .
- Show that an unbiased estimator for the population total can be found iff the first order inclusion probabilities are positive for all units in the population.
- In SRSWOR , show that E p [ s xy ] = S xy .
- Define Midzuno Sampling Design and show that this method is a PPS selection method.
- Write the estimator for population total Y under Random Group Method and show that this estimator is unbiased for Y.
- Show that the expansion estimator is equal to the population total under Balanced Systematic Sampling, in the presence of linear trend.
- Derive the approximate bias of the ratio estimator for population total.
- Show that LR is more efficient than R unless β = R .
SECTION – B
Answer any FIVE questions. Each carries EIGHT marks. (5 x 8 = 40 marks)
- Give an example to show that an estimator can be unbiased under one design but
biased under another design.
- Under any design P( . ), derive the variance of Hurwitz – Thompson estimator for population total.
- Describe the unit drawing mechanism for simple random sampling design and prove that the mechanism implements the design.
- If T( s, s′ ) is a statistic based on the sets s and s′ which are samples drawn in the first phase of randomization and the second phase of randomization respectively, then prove that
V( T( s, s′ ) ) = E1 V2 ( T( s, s′ ) ) + V1 E2 ( T( s, s′ ) ) ,
where E2 is the expectation taken after fixing the subset s and E1 is the
expectation with respect to the randomization involved in the first phase.
- Show that the estimated variance v( HT ) is non-negative under Midzuno Sampling Design for all s receiving positive probabilities.
- Show that LSS is more efficient than SRS for population with linear trend.
- Obtain Yate’s corrected estimator under LSS in the presence of linear trend to estimate population total without error.
- Describe Simmon’s unrelated randomized response model and obtain the estimate of ΠA when ΠY is unknown.
SECTION – C
Answer any TWO questions. Each carries TWENTY marks. ( 2 x 20 = 40 marks)
19 ( a ) After the decision to take a SRS has been made, it was realized that Y1 the value of unit with
label 1 would be unusually low and YN the value of unit with label N would be unusually high. In
such cases it is decided to use the estimator
if 1 s, N s
* = if 1 s, N s
otherwise,
where c is a pre-determined constant. Show that ( i ) * is unbiased for for any c.
( ii ) Derive V(* ). ( iii ) Find the value of c for which * is more efficient than . .
( 14 )
19 ( b ) State the unit drawing mechanism for Midzuno Sampling Design and show that
the mechanism implements the design. ( 6 )
20 ( a ) Derive the estimated variance of DR . ( 10 )
20 ( b ) Show that the expansion estimator is equal to the population total under
Balanced Systematic Sampling in the presence of linear trend . ( 10 )
- Derive the expressions for the approximate bias and MSE of the estimator R
and deduce their expressions under ( i ) SRSWOR, (ii) PPSWOR, and ( iii ) Midzuno Sampling.
( 20 )
22 ( a ) Show that Hansen-Hurwitz estimator dhh under double sampling is unbiased
for Y and derive its variance. ( 12 )
22 ( b ) Explain Stratified Sampling. Deduce the expressions for St , V (St ) and
v (St ) when samples are drawn independently from different strata using
( i ) SRSWOR, and ( ii ) PPSWR. ( 8 )