1 SYSTEMATIC SAMPLING

1 Systematic Sampling

1.1 Introduction
Suppose that the N units of the population are numbered 1 − N in some order. In systematic
sampling, to select a sample of n units we take a unit at k intervals from the first k units and
every kth unit thereafter. The constant k is called the sampling interval e.g. if k = 9 and the first
unit selected is 5, the sample will consist of the units numberd 5, 14, 23, 32, 41, · · · e.t.c.
In systematic sampling the selection of the first unit determine the whole sample.
The apparent advantages of this method over simple random sampling are

1. It is easier to draw a sample and often easier to execute without mistakes.

2. Intutively, systematic sampling seems likely to be more precise than simple random sam-
pling. In effect it stratifies the population into n strata, which consist of the first k units, the
second k units and so on. We might therefore expect the systematic sample to be aboout
as precise as the corresponding stratified random sample with one unit per stratum.

1.2 Description of a systematic sample:

Assume N = nk i.e N is divisible into n groups of size k. Let the units be labeled 1, 2, ..., N in
some order. The corresponding observations be ordered as

Group Observations
1 x1 x2 ··· xk
2 xk+1 xk+2 ··· x2k
3 x2k+1 x2k+2 ··· x3k
.. .. .. ..
. . . ··· .
n x(n−1)k+1 x(n−1)k+2 ··· xnk

Next select an item from group 1 at random (i.e each group with probability 1k ) suppose the
selected unit is i with the corresponding observation Xi then the systematic sample consists of
the observations xi , xi+k , xi+2k , · · · , x(n−1)k+i . If the units are each at a distance k, then this is
also called kth systematic sample. The possible number of systematic samples each of size n is
k. The possible samples together with their means are given below.

Random start Sample Composition Probability mean

1
1 1, 1 + k, · · · , 1 + jk, · · · , 1 + (n − 1)k k x̄1
1
2 2, 2 + k, · · · , 2 + jk, · · · , 2 + (n − 1)k, k x̄2
.. .. .. ..
. . . .
1
i i, i + k, · · · , i + jk, · · · , i + (n − 1)k k x̄i
.. .. .. ..
. . . .
1
k k, 2k, · · · , k + jk, · · · , nk k x̄k

1
1.3 Lahiris Method 1 SYSTEMATIC SAMPLING

Where x̄i = 1n ∑nj=1 xi j , xi j is the jth observation of the ith systematic sample. Using this
1 k
notation, it follows that X̄ = 1k ∑ki=1 x̄i = nk ∑i=1 ∑nj=1 xi j

1.2.1 Unbiasedness

Let x̄sy be the estimate of X̄ from the systematic sample, then E (x̄sy ) = X̄

E (x̄sy ) = ∑ x̄sy pr [sy]

sy

k  
1
= ∑ x̄i
i=1 k

1 k
= ∑ x̄i
k i=1

1 k n
= ∑ ∑ xi j
nk i=1 j=1

= X̄

Hence x̄sy is an unbiased estimator of X̄

Example 4.1

The possible systematic samples for N = 23 and k = 5 are

No. of systematic sample

I II III IV V
1 2 3 4 5
6 7 8 9 10
11 12 13 14 15
16 17 18 19 20
21 22 23

Since N is not in general a multiple of k i.e N 6= nk where n is the sample size, the sample
sizes may not be the same for different samples. To solve this problem Lahiri came up with the
following procedure.

1.3 Lahiris Method

Regard the N units of the population as arranged around a circle. Let k be the nearest integer to
N th
n . Select a random number between 1 − N and take every k unit thereafter.

2
1.3 Lahiris Method 1 SYSTEMATIC SAMPLING

Example 4.2

For a sample of sizen = 5, N = 23, we take k = 5 Nn = 23

5 ≈ 5 and if the random number
selected from 1 − N is 19 then our sample consists of the units numered 19, 1, 6, 11, 16

Notations

Let xi j denote the jth observation in the ith systematic sample i = 1, 2, · · · , k and j = 1, 2, · · · , n
where k is the number of samples and n the sample size of ech systematic sample. Such that
N = kn. n
∑ xi j
j=1
xi ; i = 1, 2, · · · , k denote the mean of the ith systematic sample i.e xi = n . Let N = nk i.e
the systematic samples are of equal sizes.
The variance of the mean of a systematic sample is given by

N − 1 2 k (n − 1) 2
Var (xsy ) = S − Swsy
N N
where
k n
2 1

Swsy = ∑ ∑ xi j − xi ........................................(i)
k (n − 1) i=1 j=1

is the variance among units that lie in the same systematic sample (within systematic sample)

Proof

k n
2
1
S2 = N−1 ∑ ∑ xi j − X
i=1 j=1
k n
2
1
= ∑ ∑ xi j − xi + xi − X
N−1
i=1
" j=1 #
k n 2 k n 2
1



= N−1 ∑ ∑ xi j − xi + ∑ ∑ xi − X
"i=1 j=1 i=1 j=1
#
k n 2 k 2
1



= N−1 ∑ ∑ xi j − xi + n ∑ xi − X
i=1 j=1 i=1
k n
2 k
2
⇒ (N − 1) S2 = ∑ ∑ xi j − xi + n ∑ xi − X ..................................................(ii)
i=1 j=1 i=1

From (i) we have

k n
2 k
2
2 = 1
k(n − 1)Swsy ∑ ∑ xi j − xi .............(iii) and the Var (xsy ) = k ∑ xi − X such that
i=1 j=1 i=1
k
2
kVar (xsy ) = ∑ xi − X ............(iv)
i=1
Substituting (iii) and (iv) into (ii) we have;

(N − 1) S2 = k(n − 1)Swsy
2
+ nkVar (xsy ) .........................(v)

but N = nk,so we have

NVar (xsy ) = (N − 1) S2 − k(n − 1)Swsy
2

3
1.4 Cluster and Multistage Sampling 1 SYSTEMATIC SAMPLING

(N − 1) 2 k(n − 1) 2
⇒ Var (xsy ) = S − Swsy
N N

1.4 Cluster and Multistage Sampling

A sampling procedure assumes the division of the population into a finite number of distinct
and identifiable units called the sampling units. The smallest units into which the population
can be divided are called elements of the population and groups of elements are called clusters.
When sampling the sampling unit is called cluster, the procedure of sampling is called cluster
sampling. When the list of elements (frame) is not available using an element as the sampling
unit is not feasible. The method of cluster sampling can be used in such cases e.g lists of
farms may not be available. A necessary condition for the procedure is that every unit of the
population must correspond to one and only one cluster.

1.4.1 Estimation Of Population Mean In Cluster Sampling

We assume

• Population consists of N clusters each of size M units each.

• n clusters are selected from the N clusters by SRSWOR.

Let Xi j be the survey measurements for the jth unit in the ith cluster.
X̄i = M1 ∑M j=1 Xi j
1
X̄ = NM ∑M N
j=1 ∑i=1 Xi j
X̄¯ = N1 ∑N i=1 X̄i
1

2
Si = M−1 ∑M
2
j=1 Xi j − X̄i
Sw2 = N1 ∑N i=1 Si
2
 2
S2 = 1 ∑N Ȳ − Ȳ¯
b N−1 i=1 i

Ȳ¯ = N1 ∑N
i=1 Ȳi
As the estimator of Ȳ , consider ȳ¯ = n1 ∑ni=1 ȳi . The question is, is this a good estimator?

1 n
E (ȳ¯) = ∑ E (ȳi)
n i=1

1 N
= ∑ Ȳi E(ai )
n i=1

1 N n
= ∑ Ȳi N
n i=1

1 N
= ∑ Ȳi
N i=1

= Ȳ

4
1.4 Cluster and Multistage Sampling 1 SYSTEMATIC SAMPLING

Hence ȳ¯ is an unbiased estimator of Ȳ .

So far we have considered the case when the elements of the selected cluster are enumerated.
We now consider a case when a specified number of elements is selected from each of the
selected cluster. Such a procedure is known as sub sampling procedure or two stage sampling
because the sample is chosen in two stages. The clusters that form the units of the sampling
at the first stage is called first stage units or primary sampling units i.e PSU or FSU. And the
elements or groups of elements within the clusters which form units of sampling at the second
stage are called second stage sampling units or secondary stage sampling units (SSU). The
procedure is generalised to three or more stages sand is termed as multistage sampling. We will
assume that sampling at each stage is carried out with equal probability and without replacement
i.e SRSWOR at each stage will be used.