Assigning the data into categories means going through the data case by case in a systematic way and deciding whether and how bits of data should be categorised. This requires considerable consideration in order to ensure that all the appropriate categories for all the data have been considered.

Whenever we divide data into bits in this way, meaning is lost because the data is abstracted from its context. On the other hand, the bits of data also acquire a new meaning in relation to other bits of data with which we implicitly or explicitly compare them.

The most important product of categorisation is a category set which is conceptually and empirically grounded" in the data. Categories are created, modified, divided and extended through confrontation with the data.

Choice between alternative methods of breaking up the data may be dictated by how fine-grained we want our analysis to be, and how narrowly or broadly focused our categories are.

PARADOX:
Once the data is categorised, we can examine and explore the data in our own terms. Without abstraction, comparison is not possible. But QR says, data should be analysed in context.

The ability to see the data one way, and then another is perhaps the nearest we can hope to come to coping with this paradox.

After creating and assigning categories we will refine and focus our analysis. Attention is now shifted from the "original" data itself to the data as reconceptualized through creation and assignment phases of categorisation. This shift in focus has been described as a "recontextualization" of the data. By abstracting the data from its original context, there is an obvious danger of misunderstanding or misinterpretation.

For each databit we may also hold information about the case to which it belongs, the date when it was categorised, the original context from which it comes and so on. This information may be vital to our interpretation of the data.
["Papa and journalists in New York" joke and leftist quotations from 70's are given as class examples.]

What do we gain by way of compensation? We gain the opportunity to think about our data in a new way. We can now make comparisons between all the different databits assigned to a particular category.
[Remember when we will talk about "Balanced Scorecard"]

We can compare the databits assigned to one category with those assigned to another. On this basis, we can further clarify our categories and contribute to developing the conceptual framework through which we can apprehend our data: two main tasks "splitting" and "splicing" categories.

SPLITTING: Refining categories by subcategorising data. Not all categories will require or merit subcategorisation. Even if it makes sense to subcategorise the data we have to decide whether it is worthwhile conceptually to do so. If the subcategories make sense, and seem valuable analytically, we will have to decide whether it is practically useful to subcategorise to databits.

SPLICING:   Combining categories to provide a more integrated conceptualisation. The fewer and more powerful our categories, the more intelligible and coherent our analysis. We want to concentrate our efforts on the central categories emerging from our preliminary analysis: our main interests and objectives. Likely but not necessarily; we have to avoid a mechanical approach. Issues in splicing categories:

- How central are the categories analytically?
- How are they distinguished conceptually?
- How do they interrelate?
- Are they inclusive or exclusive?
- Are they on the same status or super/subordinate?
- Does evidence of retrievals support these definitions?
- How much data do these categories encompass?
- How much overlap is there between categories?
- How do categories contribute analytically?

Back to home