Marketing research 193
Luck and Rubin (1987) define statistical analy-
sis as ‘... the refinement and manipulation of
data that prepares them for the application of
logical reference’.
After the statistical analysis stage, comes
that of interpretation – where data are trans-
formed or refined into a state which will
highlight their meaning; inductive and deduc-
tive processes are utilized.
Beveridge (1957) says that in inductive
reasoning one starts from the position of the
observed data and then proceeds to develop a
generalization that explains the observed inter-
action/situation. Deductive reasoning, on the
other hand, moves from the general to the
specific, by applying a theory to a particular
case. Data interpretation should be concluded
as objectively as possible. To ensure this, the
following points are important:
1 Honest/objective interpretations are aided by
not exaggerating or distorting the findings.
2 Interpreters should remember that a small
sample will limit the opportunity to generalize
about a large population.
3 One should not try to reach a particular
conclusion.
4 The validity and reliability of the data must be
ensured before interpreting the results, and
there should be no confusion between facts
and opinion.
Thus, the steps in the analysis of data are as
follows:
1 Put the data in order
Raw data generated by primary and secondary
research are not in a suitable state for immedi-
ate interpretation; they need to be
transformed.
Editing involves, for example, checking the
questionnaire to ensure that all the questions
have been answered and that the respondent
has given unambiguous answers. If answers
are missing or ambiguous, then steps should be
taken to either fill them in, or respondents
questioned to resolve areas of confusion.
Codinginvolves the assignation of a number,
usually to each particular response for each
question; questionnaires which are pre-coded
will save a great deal of time at this stage. Open-
ended questions also require coding, and this is
usually carried out by expert analysts, who
review a representative sample of all the
questionnaires and devise appropriate cate-
gories to which individual answers can be
assigned.
Tabulationinvolves arranging the data such that
their significance may be appreciated; data are
placed into appropriate categories which are
relevant to the research objectives. Tabulation
may be carried out manually, mechanically or
electronically. Such tables are very well suited to
variables measured by ordinal or nominal
scales, because of the limited number of
response categories. Cross-tabulation is a more
developed form of the one-way tabulation
described above, and the system allows an
investigation into the relationship between two
or more key variables by counting the number of
responses that occur in each of the categories.2 Make a survey of the data
Unprocessed data need to be transformed. The
most common way to compress data is to
calculate the data’s central tendency: mean,
median or mode. Other, more complicated,
measures of central tendency include such
measures of dispersion or range, variance and
standard deviation, and, if two or more distribu-
tion dispersions are being compared, the coeffi-
cient of variation. The results of the analysis do
not need to be presented in purely mathematical
forms; graphical display is a very useful method
of showing, for example, the frequency differ-
ences between different categories. Histograms,
line and scatter graphs and pie charts have all
been found to be better at communicating
results than bald tables of numbers.3 Select an appropriate method of analysis
If the research objectives cannot be reached by
survey and/or cross-tabulation of the data and