20.02.2013 Views

Qualitative_data_analysis

Qualitative_data_analysis

Qualitative_data_analysis

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

208 QUALITATIVE DATA ANALYSIS<br />

<strong>data</strong>. Only prejudice would prevent us from using this information if it can help us<br />

to identify patterns within the <strong>data</strong>.<br />

On the other hand, we are faced with a matrix whose meaning has been reduced<br />

to whether or not a category has been assigned to the <strong>data</strong>. The interpretation of<br />

this information depends entirely upon the decisions we have made in creating<br />

categories and assigning them to the <strong>data</strong>. We may be wary, therefore, of drawing<br />

conclusions on the basis of such limited evidence. Any inferences we can draw about<br />

regularities and variation in the <strong>data</strong> must reflect the confidence we can place in our<br />

initial conceptualizations. The matrix can be useful, but we may need to return to<br />

the original <strong>data</strong> as often as possible, for confirmation of patterns apparent within<br />

the <strong>data</strong> or to reexamine and modify our earlier judgements. With its facilities for<br />

searching and retrieving <strong>data</strong>, the computer should make it possible to do so with a<br />

minimum of fuss.<br />

If our variables describe properties of cases, we may be able to ‘recode’ these values<br />

to make a more meaningful classification. Suppose, for example, that it made sense<br />

to think of the degree to which these letters rely upon stereotypes of artistic<br />

temperament. Then we might ‘recode’ our <strong>data</strong> to discriminate between those<br />

where Woody Allen relies heavily on artistic stereotypes and those where he doesn’t<br />

use them much if at all. We might distinguish two or more values of the variable<br />

‘temperament’, such as ‘high’ and ‘low’. To recode the <strong>data</strong> in terms of these values,<br />

we would need to identify a cut-off point above which the existing values would be<br />

recoded as ‘high’ and below which they would be recoded as ‘low’. Of course we<br />

also have to decide whether to record values which are at the cut-off point itself as<br />

‘high’ or ‘low’. Deciding where to make a cut-off point is a conceptual and<br />

empirical problem. Empirically, we may be influenced largely by the pattern of<br />

existing values. Is there a natural break in the <strong>data</strong>? Does some subdivision of the<br />

range of values (e.g. midway) give a useful distinction between ‘high’ and ‘low’<br />

values? Conceptually, we have to be sure that it is meaningful to describe some<br />

values as ‘high’ and others as ‘low’. These are of course relative terms, and we have<br />

to consider the context in which they are made. If stereotypes of artistic<br />

temperament are used five times in the space of a short letter, can this reasonably<br />

described as ‘heavy’ usage?<br />

By reducing the number of values, we not only summarize the <strong>data</strong> but also<br />

render it more intelligible. We translate numbers into meaningful values. Our<br />

measurement may be less exact, but our classification is more useful. The matrix in<br />

Table 13.7 shows one possible recoding of the values in Table 13.6.<br />

We can see at a glance the patterns of usage for some of the main categories in<br />

our <strong>analysis</strong>. On the other hand, these patterns are a product of a series of decisions<br />

we have made in interpreting the <strong>data</strong> through categorizing and recoding it. We<br />

must therefore avoid any presumption that in some sense they really exist in the

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!