SYSTEMATIC

RANDOM
SAMPLING
 “…systematic sampling, either by
itself or in combination with some
other method, may be the most
widely used method of sampling.”

Levy and Lemeshow, 1999

“ A sample obtained by randomly
selecting one element from the
first k elements in the frame and
every kth element thereafter is
called a 1-in-k systematic sample
with a random start”
 In simple random sampling we
want that the samples should be
distributed randomly.
 In reality the random selection
may be like this:
 In systematic sampling we force
to select samples “evenly” from
the list (sampling frame), and
only the first unit is selected at
random.
 The rest being selected according
to a predetermined pattern.
 to select a systematic sample of n
units, the first unit is selected
with a random start r from 1 to k
sample, where k=N/n sample
intervals, and after the selection
of first sample, every kth unit is
included where 1≤ r ≤ k.
 Let N=100, n=10, then k=100/10.
 Then the random start r is selected
between 1 and 10 (say, r=7).
 So, the sample will be selected from
the population with serial indexes of:
 7, 17,27,.........,97

i.e., r, r+k, r+2k,......., r+(n-1)k

[What could be done if k=N/n is not

an integer? ]
 Consider, n=175 and N=1000. So,
k=1000/175 = 5.71
 One of the solution is to make k rounded
to an integer, i.e., k=5 or k=6.
 Now, if k=5, then n=1000/5=200; or,
If k=6, then n=1000/6 = 166.67 ~
167.

Which n should be chosen?

 if k=5 is considered, stop the
selection of samples when n=175
achieved.
 if k=6 is considered, treat the
sampling frame as a circular list and
continue the selection of samples
from the beginning of the list after
exhausting the list during the first
cycle.
 An alternative procedure is to keep k non-
integer and continue the sample selection
as follows:
 Let us consider, k=5.71, and r=4.
 So, the first sample is 4th in the list. The
second = (4+5.71) =9.71 ~9th in the list,
the third =(4+2*5.71) =15.42 ~ 15th in the
list, and so on. (The last sample is:
4+5.71*(175-1) = 997.54 ~ 997th in the
list).
 Note that, k is switching between 5 and 6.
 Simple to implement
 Saves time and cost
 It's variances are most often smaller
than other alternative sampling
technique, when it is suitable to use.
 It can interact with a hidden periodic
trait within a population hence the
technique will be compromised since
there will not be random and
representativeness of the sample
 May be started without a complete
listing frame (say, interview of every
9th patient coming to a clinic).
 There is the possibility of losing vital
information from the population.
 It may not be possible to select the
required sample size if the population
is too small.
 It may not be good for periodic data.
 The process of selection can interact
with a hidden periodic trait within the
population.
 In this case, the end of list is connected
to the beginning of the list, making the
list circular. This allows the random start
r to start between 1 to n (1<r<n), rather
than between 1 to k as in the linear
systematic sampling.
 Say we want to take a sample of size 10
from a population of 100.
 We will select the first sample randomly,
say, 85th element.
 So, our sample will consist of the
following elements:

85, 95, 5, 15, 25, 35, 45, 55, 65, 75

 Even preferred over SRS
 When no list of population exists
 When the list is roughly of random
order
 Small area/population
 Systematic sampling is easier to perform in the
field and hence is less subject to selection
errors by fieldworkers than are either simple
random samples or stratified random samples,
especially if a good frame is not available.
 Systematic sampling can provide greater
information per unit cost than simple random
sampling can for populations with certain
patterns in arrangement of elements.
 Provides a better random distribution than SRS
 Estimator of the population mean:

1 n
̂  y sy   yi
n i 1
 Estimated variance of mean:
n
 1 
 
2
 
V̂ y sy
 ns
 1   where s2  
n  1
 i y  y sy
2
 Nn  i 1

 Bound on the error of estimation:

2
   n s “Assuming a
2 Vˆ y sy  2 1   randomly ordered
 Nn population”
 The estimated variance of y sy is identical to
estimated variance of y obtained by using
SRS.
 This result does not imply that the true
variance of y sy is the same as that of y
 Discussion:
SRS SYSTEMATIC
2
 2

V̂ y 
n
1    
V̂ y sy 

n
1  n  1 
n  N

ρ is a measure of the correlation between pairs

of elements within the same systematic sample
 Discussion:
ρ close to 1 : Systematic sample has higher
variance of sample mean
ρ is negative : Systematic sampling more precise
than SRS
ρ close to 0 and N : Systematic sampling is equivalent to
fairly large SRS

 When systematic sampling is nearly equivalent

to srs, we can estimate V y sy  by the estimated
variance from srs.
 In addition, these approximation works well for
a systematic sample for a random population
behaves.
 Random population elements

A population is random if the elements of the

population are in random order.
 Ordered population elements

A population is ordered if the elements of the

population are have values that trend upward
or downward when they are listed
 Periodic population elements

A population is periodic if the elements of the

population are have values that trend upward or
downward when they are listed tend to cycle upward
and downward in a regular pattern when listed.
 Estimator of the population total:
N n
Tˆ  Ny sy   yi
n i 1

 Estimated variance of total:

 
n
   n 
2
V̂ T̂  V̂ Ny sy  2 ns
 N 1  s 2  1   y i  y sy
 N  i 1
2

 Nn

 Bound on the error of estimation:

2
2 V Ny sy   2 N 1  
“Assuming a
ˆ 2 n s
randomly ordered
 Nn population”
 Estimator of the population proportion:
1 n
p̂sy  y sy   yi
n i 1

 Estimated variance of total:

n  p̂sy q̂sy
V̂ p̂sy   1  

where q̂sy  1  p̂sy
 N  n 1

 Bound on the error of estimation:

n  p̂sy q̂sy
2 V̂ p̂sy   2 1  
 “Assuming a
 N  n 1 randomly ordered
population”
 Sample size (means):

N 2 B2
n where D 
N  1D   2 4

 Sample size (totals)

N 2 B2
n where D 
N  1D   2
4N 2
 Sample size:
Npq
n
N  1D  pq

where
B2
q  1  p and D 
4