The Marketing Book 5th Edition

192 The Marketing Book

method, it is quick to use and relatively simple
to administrate – it does not require a sampling
frame. However, there is the possibility that the
interviewer shows bias in the way the individ-
ual units are selected and in the difficulty that
may arise in uncovering relevant and available
quota controls.

Probability versus non-probability

sampling techniques

Tull and Hawkins (1993) provide the following
list of factors that are worthy of consideration
when choosing a sampling method:

1 Are proportions and/or averages required or
are projectable totals needed?
2 Are highly accurate estimations of population
values necessary?
3 How large might non-sampling error be? What
size of error due to frame choice,
non-response, measurement and population
specifications is likely?
4 Will the population be homogeneous or
heterogeneous with respect to the
characteristic of interest?
5 What will be the cost if the results are
above/below the required error tolerance?

They say that ‘... the need for projectable
totals, low allowable errors, high population
heterogeneity, small non-sampling errors and
high expected costs of error favour the use of
probability sampling’.

Size of sample

The size of a sample depends, in the main,
upon the required degree of accuracy that the
research objectives demand. This will depend
upon:

1 The degree of variability in the population –
the more heterogeneous that population, the
larger the sample size required.
2 The presence of population subgroups – the
sample must be large enough to allow for a
valid analysis of these.

Sample size estimation depends upon:

 Judgement– rests on the experience of the

research consultant. But research workers

should beware of making an arbitrary choice,

ignoring such factors as cost, value and the

required level of accuracy – this is the method

of last resort.

 What can be afforded– though commonly

used, this method ignores the value of the

information to be collected, only looking at

the cost. For example, a small sample may be

more useful, though of a higher cost per unit,

than a larger sample if the collected

information is of a high value.

 Required size per cell– used in quota and

stratified sampling techniques. It is usual to

accept, as a minimum, 30 units per cell before

any statistical analysis can proceed. Thus, if

there are two age groups and five geographical

areas to be sampled, 10 cells will result, hence

a sample of 300 units is required.

 Statistical methods– sample sizes may be

calculated using the formula:



N

=

Required level of accuracy

Level of confidence

where = standard deviation and N= size of

sample.

Analysis of the results

The researcher will now be in possession of

data from both primary and secondary sources;

it must now be processed such that it is possible

to draw appropriate conclusions.

In commencing the analysis, two questions

need to be answered:

 With reference to the research objectives,

what meanings should be obtained?

 What statistical methods should be employed

to obtain those meanings given the way in

which the data were collected?