Chapter Summary

A key message of this chapter is that no single approach to data analysis is suitable for all occasions or for all research objectives. There are some contexts in which variable-based analyses may be better and others where quantitative case-based approaches are more suitable. Researchers have to make their own decisions, preferably based on a firm understanding of the advantages and limitations of the two main approaches. Thus variable-based analyses facilitate data reduction, they can make precise comparisons of differences between groups of cases, they can accurately measure covariation, they can establish the relative importance of independent variables, and they are good for exploration and verification. In contrast, case-based analyses keep the focus on cases, they can handle small-n research where the number of cases available is limited, they facilitate causal analysis, counterfactual assumptions can be controlled, and they are good at producing recipes for achieving a desired outcome.

At the same time, researchers need to bear in mind that variable-based analyses need many cases, that those cases tend to become invisible once the analysis begins, that asymmetrical patterns are unseen, that regression-based techniques are restricted to linear patterns, that limited diversity gets overlooked, that there is too much focus on statistical inference, and that such analyses are not good at establishing causality. On the other hand, quantitative case-based analyses can be used only when a number of conditions are met, solutions are, furthermore, very sensitive to the decisions and assumptions made by the researcher, they are not well suited either to exploration or verification, and there are dangers in the solutions of triviality, irrelevance or contradiction.

Researchers can, however, mix these methods and mix their data in various ways. Methods may be phased in different sequences, or they may be concurrent. There may be differences in emphasis. Mixing data involves transforming data in one dataset in various ways so that they can be incorporated into another.