Designs

32 Sample Survey

Thus, irrigated ficlds

1N N P= proportion of

and
E -N= NP
estimating a population proportion becomes that
Thus, the problem of variate as above. Ir
mean by defining the characteristic, the
of estimating a population
sample of size n possess that
n units out of a random Ths
sample proportion isgiven by p m/n.

Hence, an unbiased estimator of P is given by

P= (n/n) =p (2.4.1)
given by
THEOR FM 2.4.1 In _amplihg,wor, the variance ofp is
NPO|n (N- 1) (2.4.2)
Vp) = (N-n) POln (N- 1) = (|-)
where = I-P. The proof is obvious.
COROLLARY I In sampling, wr, the variance of p is given by
V(p) = PO/n (2.4.3)
COROLLARY 2 The variance of G= Np, the estimated total number
of units with some desired characteristic, is given by
vÔÑ) = N² (N-n) POln (N- 1) (2.4.4)
THEOREM 2.4.2 In sampling, wor, an unbiased estimator of V(p) is
given by
op) = (1-f) pgln--1) (2.4.5)
The proof is obvious.
COROLLARY 1 In sampling, wr, an unbiased estimator of Vp) is given
by
p) = Pg/(n-1) (2.4.6)
COROLLARY 2 An unbiased estimate of the variance of N = Np is
given by
o(Ñi) = N(N-n)pg/(n-)= (1-f) N² pq/(n-1) (2.4.7)
 Basic Methods of Simple Random Sampling 33
COROLLARY 3 The coefficient of variation of p is given by
CV = [PQlnj2/P = (Q/nP]2 (2.4.8)
ExA MPLE 2.4 A list of 3000 voters of a ward in a city was examined
for measuring the accuracy of age of individuals. A random sample of
300 names was taken, which revealed that 51 citizens were
shown with
wrong ages. Estimate the total number of voters having a wrong des
cription of age in the list and estimate the standard error.
Here, N= 3000, n=300, nË == 51, p = 0.17
The estimate of the total number of voters
of age is obtained by
having a wrong description
NË = Nn= (3000) (0.17) = 510
(i) If sampling, wr, is considered, the estimate of the
is given by standard error
= N (pqln- 1)]2
=3000[0.17) (0.83)/50j112
= 159.3
(ii) If sampling, wor, is
given by
considered, the estimated standard error is
SR, = NI(1-) pg/(n-1)]12
3000[(1 0.10) (0.17) (0.83)/50]/2
151.1

2.5 COMBINATION OF
UNBIASED ESTIMATORS
There are situations where the estimates based on
have to be pooled to get a combined estimate. Ifseveral samples will
t, (i = 1,2,.. ., m)
are unbiased estimators of a parameter 8,
dent. then the pooled estimator which are mutually indepen
m

2tlm (2.5.1)
is also an unbiased estimator of 8. The variance of i is given by

V) =2 VNm (2.5.2)
 34 Samole Survey Designs

and estimate of variance

-DPmom-1) (2.5.3)
M+) =
consider the problem for the following cases:
We shall sampling for variables
(i) Simple random
Simple random sampling for attributes
(i)
Simple Random Sampling for Variables
2.5.1
means, each of
suppose that p1, P2,. ..,Pm are the sample
Let us sizes n1, n2, . ., Nm. A pooled
which is drawn independently, with
estimator of all samples is given by taking
(i) tbe arithmetic mean of m estimates

j'= ilm (2.5.4)

i

(ii) the weighted mean of m estimates

(2.5.5)

n=
where

When Sampling is With Replacement Method

The sampling variances of f' and y" are given by
g2 m
Vã)= 2 VP)/m² m²
(2.5.6)

g2
and (2.5.7)

Unbiased estimators of V(P) and V) are provided by

o(') =E9-Fm(m-1) (2.5.8)

and
m n

o(ù") -Æ u-n(n-1) (2.5.9)

 Bas1c Methods of Simple Random Sampling 35

Itshould be noted that the estimate given in relation (2.5.5) is more

cficient than that given in relation (2.5.4) and can be verified by com
paring the variances in both the cases.

When Sampling is Without Replacement Method

The sampling variances of f' and ù' are given by
m

Vig) = E V()/n i(1-)Sim²n, (2.5.10)

and

V(ù) = (2.5.11)

Unbiased estimators of V(ù) and V(P) are provided by

m

(2.5.12)
and
m

(2.5.13)
The estimators y' and y'' come identical if the sample sizes n), n2,...
nm are the same.

2.5.2 Simple Random Sampling for Attributes

The results obtained in the previous section can be applied to
random sampling for attributes also. If p1, pz, . .., pm are the simple
proportions based on n samples of size ni, n2 , ..., nm each of sample
which
is drawn independently. A pooled estimator of all
samples is given by
taking
(i) the arithmetic mean of m estimates
m

(2.5.14)
(ii) The weighted mean of m estimates

p =np1n (2.5.15)

where
 Survey Designs
36 Sample
ReplacementMethod
is Wih
When Sampling P'andp" are given by
variances of
The sampling
(2.5.1e
and
P(l - P))n (2.5.1
I ( ) E V(p)/n² =
by
of Vp) and Vp") are provided
Unbiased estimators
(P)= (pr
-
p/m (n - 1) (2.5.1

and

(P) = np: (1 - p)ln² (n, - 1) (2.5.1

Another estimator is given by

P)= p (1 - P)/n (2.5.20

Method
When Sampling is Without Replacement
The sampling variances of p' and are given by
VP) = / (N- n) P(I P)/n, (N- D² (2.5.21

and

VP") = }) (N-n) P(| - P)n, (N - I) (2.5.22

Unbiased estimators of V(P') and V(P") are provided by

P') -}(p, - Pm (m - 1) (2.5.2
and

oP") - E(N- m) nË p. (1 - p)/n² N(nm - 1) (2.5.2

 Sirnple Random Sampling 37
Basic Me thods of

2.6 CONFIDENCE LIMITS

parameter, it becomes neces
After having the est1mate of an unknown
estimates and to construct somc
sary to measure the reliability of these we assume that
confidence limits with a given degree of confidence. If
population mean I,
the estimator p is normally distributed about the given
lower and upper confidence limits for the population mean Y are
by
YL = ù- tay n-) s [(I -f)in]2 (2.6.1)
and

Yu = ù + tiy n-) s [(1 -fjn]? (2.6.2)

where tas n-) Stands for the value of student's twith (n - 1) degrees
of freedom at a level of significance. Similarly, the confidence limits
for the population total may be written as
YL = Nù - t(zs n-1) Ns ((1-)ln]n (2.6.3)
and

Yu =Nù + (ay n-l) Ns[(1 pln2 (2.6.4)

EXAMPLE 2.5 Signatures to a petition were collected on 700 sheets.
Each sheet was provided with space for 50 signatures, but the sig
natories put their signatures in erratic ways and the number of signa
tures per sheet was not definite. 12 sheets were spoiled in transit. Of the
available sheets, a random sample of S0 was drawn and the numbers
of signatures per sheet counted, which are shown as below:
Number of signatures () .. 52 S1 46 42 40 37 32 29 27 15 14 10 8
) Number of sheets (n,) 1 2 21 8 7 2 2 1 2 1 1 1

Estimate the total number of signatures to the petition and calculate

95% confidence limits.
We have
3)
N= 700 - 12= 688, n = ) n, = 50
Eny,=1992, .vi-84820
Therefore, the estimate of total signatures is given by
1)
Y= Nù = (688) (1992) 50 27,410