Transcript File

DETERMINING SAMPLES FOR A
SAMPLE SURVEY
When a sample survey is to be completed, care must be taken when determining the
sample that comes from the population.
How the individuals are chosen may impact on the reliability of the sample survey.
To have a reliable survey, the sample must be representative of the population. If it is not
representative the survey will contain a bias and may be called into question.
The number of individuals chosen can also impact on the reliability of the sample survey.
One would think that the greater the sample size the more reliable the survey. That is
essentially true, but there is very little to be gained by increasing the sample size past a
certain point.
There are 3 methods that can be used to achieve a representative sample (non-biased).
1. Random Sampling
Individuals are chosen in a haphazard manner from population. No
consideration is given to declare an individual within a population
as unsuitable. Any population individual can be chosen as part of
the sample.
2. Stratified Sampling
Planning is used to carefully select certain individuals so that the
characteristics of the sample match the characteristics of the
population. Gender, age and ethnicity are often taken into account
in the sample selection.
3. Systematic Sampling
This involves using some system to select individuals from a list.
Selecting every tenth or twentieth person from a list so that each
individual chosen is taken based on some arbitrary rule and not
some planned criteria.
Each of the above methods should produce a representative sample that is necessary to
produce a reliable survey.
Another requirement to produce a reliable survey is to have a sufficiently large enough
sample.
The larger the sample the more reliable the survey.
Reliability can be quantified using the term ‘margin of error’ (ME). The greater the margin
of error, the less reliable the survey.
Margin of error is expressed as a percentage and an indication as to how much the
samples numbers can be off by.
There is actually a formula that allows us to link the reliability of the sample to the
sample size. It will produce reliable results 19 times out of 20 within the margin of error
desired given a certain number of individuals (n) sampled.
0.9604
n
ME 2
n = number of data values
ME = margin of error transformed
from percentage
What sample size is required to produce a margin of error of 1%.
0.9604
n
ME 2
0.9604
n 
(0.01)2
0.9604
n 
0.0001
n  9604
You need 9604 in your sample for this degree of accuracy.
ME = 1% = 0.01
The formula can be used the other way around. If we know how many people are in our
sample, we can determine our margin of error.
If 1000 customers are involved in a sample survey, what is the margin of error for that
survey?
0.9604
n = 1000
n
ME 2
0.9604
1000 
ME 2
1000ME 2  0.9604
0.9604
ME 
1000
ME 2  0.0009604
2
ME  0.0009604
ME  0.03099
Convert 0.03099 to a percentage.
0.03099  100 = 3.099% ≈ 3.1%