you already know, as a jumping-off point for getting into the data, is entirely

legitimate: it will not contaminate the objectivity of your overall analysis.

So, in sum, it is entirely legitimate, when ¬rst getting to grips with the data, to

start applying a simple ˜common sense™ approach to the interpretation of the data.

Beginning to identify and understand the main patterns

and relationships

Before beginning the process of analysing the data in close detail “ remembering

the Hermeneutic Circle principle “ it is worthwhile, while in a preliminary

overview analysis mode, to look for any broad patterns and relationships that

are emerging from the data. Here, it is helpful to look at how, for example, any

attitudinal differences may, or may not, correspond with behavioural differences.

Now is the time to also examine whether there are similarities where you would

have expected to ¬nd differences, and so on. All of this will be helpful in

beginning to compare and contrast how different competing putative storylines

in the data may be building up a picture in your own mind about what this data

is saying. You may wish to begin to follow up on any aspects of the data that you

feel are unlikely to be accurate, by searching out any corroboratory evidence. At

this point, you will also be able to make an initial assessment of whether or not

your data is going to be able to meet the initial research objective (possibly by

setting in train calls for extra, more detailed analysis to plug any gaps).

Establishing what the total sample story is telling you

Having understood the overall relationships at work within the data set, it is now

time to start laying down the fundamental story that is being told by the data.

As a general rule, the ˜story™ behind most survey data can be told round the

total sample (assuming it is representative). So, when beginning to analyse any

data set, it is helpful to quickly move to a position where you have established

the storyline for the total sample. That is, at this stage, do not worry about

analysing variations within the different analysis sub-groups. Simply concentrate

on studying your total sample data in relation to your overall analysis objectives.

Then, see if you can tell yourself (in ¬ve minutes) the overall storyline for the

total sample that is emerging about the problem under investigation.

A helpful tip here, for a survey research study, is to write on the questionnaire

the results for the ˜total sample™ (without worrying at this point about any sub-

group variation). Then, working with a photocopy of the questionnaire (showing

the total sample), ˜cut and paste™ (group together) the questions that seem to

answer the various analysis objectives. In this way, you are beginning to create

a ˜story™ that could form the running order of your ¬nal presentation or report.

This is clearly preferable to the rather na¨ve approach of simply reporting the

±

answers to all of the questions, in the order of the questionnaire.

88 Developing the analysis strategy

Re¬ne the main storyline with the sub-group analysis

Having understood the overall total sample storyline, it is now time to begin

to understand what is happening in the sub-groups. Most sub-group analysis

tends only to re¬ne the overall storyline. Rarely are the sub-groups suppressing

a totally different storyline that needs telling for different sub-groups. Most sub-

group analysis simply adds a variation to the overall theme. Rarely does this

overall analysis lead to the necessity of telling a series of multiple sub-plot stories

• There is a statistical technique called Chaid (Chi-squared Automatic Interaction

Detector).

• This technique is helpful in ensuring the analyst does not fail to detect the key

determinants of attitude or behaviour by not properly structuring the sub-group

analysis. Thus, it could be that an analysis of the total sample by a separate cross-

break for age, and a separate analysis break for gender, may fail to unearth

differences of, say, age within gender, that are critical to telling the true story.

• The Chaid technique, by systematically examining the 'distance' between every

possible combination of sub-group, will alert the analyst to where they are missing

key insights

• Chaid takes a given dependent variable (the one you want to be able to predict), and

a set of independent variables (the ones you want to use to predict the dependent

variable), and develops a flow chart according to how strongly different independent

variables seem to explain differences in the dependent variable.

• For example, if we wish to examine the variations in the number of fiction books read

each year by the total population by age and gender sub-group differences, the final

outcome of the Chaid analysis might look as follows:

Total Sample

Average no. 5 books

Men Women

(3 books) (7 books)

Under 34 Over 34

Under 34 Over 34

(3 books) (11 books)

(2 books) (4 books)

ABC1 C2DE

(18 books) (4 books)

Thus, the storyline here is that (older) upmarket women are our big fiction readers.

Figure 7.1 “ The Chaid technique for analysing sub-groups.

89

Holistic data analysis as successive waves of analysis at varying levels

that are at complete odds with the total sample. (This can, of course, happen,

but not in the majority of scenarios.)

Here, with quantitative data, a useful rule of thumb is to quickly establish what

differences will be needed between, for example, two key sub-samples, for there

to be any statistically signi¬cant variation, and then to scan the data only looking

at data that is distanced by the requisite percentage points, or more. So, if you

had two key sub-samples of 200 each, you would go looking for differences

of at least 10% points, or more. (In the Notes section we identify a source for

checking the signi¬cance of the difference between two sample statistics. And in

Figure 7.1, we summarize a technique for analysing sub-groups.)

Holistic data analysis as successive waves of analysis

at varying levels

So, to conclude this chapter on the general principles of holistic data analysis, a

picture is emerging of the holistic analyst being involved in successive waves of

analysis. The ¬rst wave will concentrate on establishing the fundamental patterns

and shapes emerging from examining the total sample in fairly broad strokes.

The next wave of analysis will put the spotlight on the detail. Speci¬cally,

Wave 1:

Identify the fundamental

structure of, and

relationships within the

data (examine the

relationships between

Wave 6:

attitude and behaviour, Wave 2:

Test final outcome of

and the data in relation Think 'total sample'

analysis against your

to the business only: develop competing

original research and

working narratives and

objectives).

business objectives.

test against business

objectives.

Wave 3: