194 The Marketing Book
more sophisticated methods of analysis are
required, then consideration should be given as
to which particular analytical techniques will
provide the appropriate information. However,
there is a vast array of available techniques, so
some thought needs to be given to the way in
which the most appropriate method(s) is (are)
selected.
Luck and Rubin (1987) offer the following
scheme:
1 What is the technique required to show?
A common request is to show whether the
results are significant, i.e. are there
significant differences between various
groups or could the results have occurred
by chance because only a sample of the
population was under investigation, not the
entire population.
2 What scale was used to measure variables?
Only certain arithmetic manipulations are
allowed on certain types of scale – it
depends upon what level of measurement
was reached.
Non-metric scales, where the data were
qualitative rather than quantitative, include
nominal and ordinal scales.
Metric scales work in real number systems
and include ratio and interval scales.
3 Parametric and non-parametric data.
Parametric data are those which are
distributed around a mean/central value in
a symmetrical manner, as in a normal
distribution, and have been collected, at
least, using an interval scale and may be
analysed using probabilistic tests of
statistical significance.
Non-parametric data have a distribution
profile which does not conform to the
normal curve of probability and
appropriate tests assume that the variables
have been measured using nominal or
ordinal scales.
4 Number of variables to be analysed
Univariate analysis: where a single variable is
analysed in isolation.
Bivariate analysis: occurs where some form of
association is measured between two variables
simultaneously.
Multivariate analysis: this investigates the simul-
taneous relationships between three or more
variables.5 Dependence and independence
Analysis may involve an investigation into the
relationship between variables. By relation-
ships is meant that changes in two or more
variables are associated with each other. It may
be important to be able to calculate by how
much the independent variables are respons-
ible for variations in the dependent variable.6 How many samples are involved?
The choice of an appropriate statistical test
depends upon whether the data are being
tested to measure (a) the significant differences
between one sample and a nominated popula-
tion, (b) the significant differences between two
related or independent samples, (c) the sig-
nificant differences amongst three related or
independent groups, or (d) correlation and
their significance tests.Presentation of the final report
Research reports should say what they are
supposed to say and do so in a style appro-
priate to the intended readership. Those which
are intended for a technically educated reader-
ship may be written using specialist terms and
may discuss, in detail, the complex issues of the
research process. Those reports intended for a
more general readership should not be used as
an opportunity to impress with an overt dis-
play of technical language/jargon and subject
matter. General readers are interested in the
results, not in the way they were reached,
though, of course, the appropriate amount of
background to the project will need to be
provided.