Sampling Methodology

Lecture 2: Simple Random Sampling (SRS)

Date: 19 May 2023

Course Teacher:
Mohammed Nazmul Huq
 Definition of SRS

Simple Random Sampling (SRS) is a method of selecting the units

(samples) from population where each population unit has an equal
chance of being selected in the sample

• Simple random sampling forms the basis of most of the sampling

designs

• Simple random sampling is always an Equal Probability Sampling (EPS)

design
• SRS is more suitable when population is small, homogeneous & readily
available
 Examples of SRS

• Daraz is planning to identify the satisfaction level of their customers

in Savar on the online shopping. The authority has a list of 1,500
customers. A SRS of customers may be taken to know the satisfaction
level.
• The EMCH authority is interested to know the level of satisfaction of
their patients admitted over the last month. The inpatient register
has a record of 900 patients and a SRS of patients may be drawn to
evaluate patient’s satisfaction.

• bKash wants to know the amount (in TK) of transactions per agent per
day in Savar area. It has a list of 250 agents in Savar. A SRS of
agents may be selected to estimate the transaction amount.
 Replacement of Selected Sampling Units

Sampling schemes may be without replacement

('WOR') or with replacement (WR)

• Simple Random Sampling with replacement

(SRSWR): In SRSWR, the units are drawn one by
one from the population and each selected unit is
replaced back to the population before executing
the next draw.

• Simple Random Sampling without replacement

(SRSWOR): In SRSWOR, the units are drawn one
by one from the population and each selected unit
is not replaced back to the population before
executing the next draw.
 Replacement of Selected Sampling Units

Sampling without replacement (WOR) or with replacement (WR)

For example:

• If we catch fish, measure them, and immediately return them to the

water before continuing with the sample, this is a WR design. We might
end up catching and measuring the same fish more than once.

• However, if we do not return the fish to the water (e.g. if we eat the
fish), this becomes a WOR design.

Obviously, in practice there is no reason of interviewing the same

individual twice!
 Drawing a SRS

To ensure randomness in the

selection, the method of selection
must be independent of human
judgment. There are two basic
procedures:

• Lottery method

• Random numbers method

Examples……
 Steps in drawing a SRS

• Assign serial numbers to the units in the population from 1 through N

• Decide on the random number table to be used
• Chose a random number, which is ≤ N
• Move on the next random number not exceeding N
• If any random number if repeated, discard it
• The process stops as soon as the desired sample size is achieved

Computer generated random numbers can also be used.

MS Excel command: =RANDBETWEEN(1, N)
 Drawing a SRS - Example
Random numbers
05 74
12 16
20 03
10 07
10 06

• Select 5 students from the first page of the attendance register

• Who are selected?
• Find at least two problems with this process
 Number of possible samples
• In SRSWR:
Number of possible samples = N n

• In SRSWOR:
Number of possible sample = N Excel command:
N
Cn =  
n =COMBIN(N,n)

Here,
N = population size
n = sample size
 Estimation of population characteristics in SRS
Estimating population mean, total, variance and standard deviation
(for SRSWOR)
n

x
• Sample mean:
i
Xˆ = x = i =1

n
n
N
• Total: ˆ
X = Nx =
n
xi =1
i

( x − x )
2
i

• Sample variance: v ( x) = s =
2 i =1

n −1

• Sample standard deviation: Positive square root of the sample variance, i.e. s2
 Estimation of population characteristics in SRS
Variance and Standard Error of sample mean (for SRSWOR)

s2  n 
• Variance of sample mean: v(x ) = 1 − 
n  N

• Standard Error of mean: SE ( x ) = v ( x )

here,
f=n/N is called sampling fraction
(1-f) is called finite population correction (fpc)
 Estimation of population characteristics in SRS

Relative error of sample mean (for SRSWOR)

SE ( x )
• Coefficient of variation (CV): CV ( x ) =
x

Note:
• Standard error of mean provides the absolute error
• For making comparisons, relative measures are used
 Estimation of population characteristics in SRS
Example: Suppose you want to open a ‘fast food’ shop at Jamuna Future
Park. There are 250 fast food shops currently operating in the mall. As
a beginner, an idea about monthly sell of the existing shops would be
useful. Under this circumstance, you select 10 shops randomly and the
total sells (in ‘000 TK) over the last month of the sampled shops are as
follows:
75 95 105 99 110 97 85 90 100 101

• Find an estimate of the monthly average sells of the shops

• Also estimate the standard error and relative errors of the sample
mean
 Estimation of population characteristics in SRS
Solution: n

• Estimate of the monthly sells per shop:

x i
Xˆ = x =
957
i =1
= = 95.7
n 10
The average monthly sells of fast food shops at Jamuna Future Park is TK 95,700

• Standard error of the sample mean: SE ( x ) =

s2 
1 −
n 
=
102.9 
1 −
10 
 = 3.14
n  N  10  250 

The average monthly sells per shop may vary by TK 3,140

SE ( x ) 3.14
• Relative errors of the sample mean: CV ( x ) = = = 0.033
x 95.7

The average monthly sells may vary 3.3%

 Estimation of population characteristics in SRS
MS Excel Command:

• Estimate of the sample mean: = AVERAGE(number1, number 2, ……)

• Sample variance: = VAR.S(number1, number 2, ……)

• Sample standard deviation: = STDEV.S(number1, number 2, ……)

 Estimation of population characteristics in SRS
Estimating population proportion (for SRSWOR)
r
• Estimation of population proportion: Pˆ = p =
n

• Sample variance of P: v ( p) =
p (1 − p )  n 
 1 − 
n −1  N

• Standard error: SE ( p ) = v ( p)

here,
p = sample proportion
r = number of cases posses the attribute of interest
n = sample size
 Estimation of population characteristics in SRS
Example: Suppose a social scientist is interested to know the proportion
of women watch Indian channels. She selects 1,500 women from Mirpur
area randomly and finds that 975 regularly watch the Indian channels.
Suppose the total number of women residing in Mirpur area is 1 million.

• Estimation of population proportion: Pˆ = p =

975
= 0.65
1500

0.65(1 − 0.65)  1500 

• Sample variance of P: v ( p) =
1500 − 1
 1 −  = .00015
 1000000 

• Standard error: SE ( p ) = .00015 = 0.012

 Determination of Sample Size
Sample size for estimation of population mean (continuous data)

z 2 2
Formula 1: n0 =
d2

Formula 2: n0
n =
n0
1+
N
here,

n0 = estimated sample size;

z = statistical certainty chosen (1.96 for 5% level of significance);

σ = population standard deviation, obtained from prior knowledge of the population;

d = width of the interval desired (precision);

n = desired sample size.

 Sample size for estimation of population mean
Example: We want to estimate the mean mobile bill per month of the 35
WMASDS-students. We wish to estimate the true mean to within 10TK
with 95% confidence. The standard deviation of monthly mobile bill is
around 50TK. What is the required sample size ?

z 2 2 1.962  502
Formula 1: n0 =
d 2
=
10 2
= 96.04

n0 96.04
n= = = 25.65
Formula 2: n0 96.04
1+ 1+
N 35

The minimum sample size required is round up to 26 students

 Determination of Sample Size
Sample size for estimation of population proportion

z 2 pq
Formula 1: n0 =
d2

n0
Formula 2: n =
n
1+ 0
N

here,
n0 = estimated sample size;
z = statistical certainty chosen (1.96 for 5 percent level of significance);
p = estimated prevalence; (0.5 if unknown); q = 1-p;
d = precision desired (usually consider 0.05);
n = desired sample size.
 Sample size for estimation of population proportion
Example: To estimate the proportion of 30 Male students who smoke,
what sample size is required to achieve a 95% confidence interval of
width ±5% (that is to be within 5% of the true value)? A study some
years ago found that approximately 30% of the students were smoker.

z 2 pq 1.96  0.3 (1 − 0.3)

2

Formula 1: n0 = 2
= 2
= 322.69
d 0.05

322.69
Formula 2: n=
322.69
= 27.45
1+
30

The sample of 27 male students is required

 Determination of Sample Size
Sample size for impact evaluations (for proportions)
2
 Z1− 2 p(1 − p) + Z1−  p1 (1 − p1 ) + p2 (1 − p2 ) 
Formula : n =  D
( p2 − p1 )
2

here,
n = estimated sample size;
D = design effect;
p1 = estimated prevalence at the time of the first survey;
p2 = expected prevalence to be achieved;
p = (P1 + P2)/2;
Z1-α = z-score corresponding to desired level of significance;
Z1-β = z-score corresponding to the desired level of power.
 Determination of Sample Size
Sample size for impact evaluations (for proportions)

The size of the sample usually depends on:

a. the number of groups of interest being studied

b. the amount of change expected in the indicator

c. the level of confidence needed for final conclusions, and

d. the probability of detecting a difference in the indicator when one

actually exists
 Advantages and Disadvantages of SRS
Advantages:
• Easy
• Provides basis to compare with other methods
• Provides foundation for much of statistical theory

Disadvantages:
• All population units must be identified
• Sampled units may be widely dispersed
• Certain sub-groups may be completely overlooked
• Time consuming and costly
 Reference Book
• Singh Ravindra and Mangat N S (1996): Elements of Survey
Sampling, Vol 15, Kluwer Academic Publishers, The Netherlands.

• Islam M Nurul (2005): An Introduction to Sampling Methods –

Theory and Applications, Book World, Dhaka.

• Cochran, W. G. (2002): Sampling Techniques, 4th Edition, Wiley

Eastern, New Delhi.

• Thompson, S. K. (2012): Sampling, 3rd Edition, John Wiley and Sons

Inc., New York.