8.2.3 Designing the case study

Unit of analysis

The first decision was to define the unit of analysis - the level of the case. In classic case study research a single individual is the object of study (and thus the unit of analysis). More common nowadays is case study research that defines organisations or programs as the basic unit. Some of the candidate academic libraries were responsible for multiple electronic scholarly publishing projects. It was decided to select a specific project (usually the lead project) within a library as the basic unit. This is because the libraries are responsible for a wide range of activities and trying to cover all of these would dilute the focus on the electronic publishing projects.

Project selection

A case study may be about single or multiple cases. It is possible to generalise from single cases (in some analytic way) but multiple-case studies can strengthen or broaden such generalisations (similar to the advantages of multiple experiments). [Yin, 1998] distinguishes between literal replication (where the cases are designed to corroborate each other) and theoretical replication (where the cases are designed to cover different theoretical conditions). In the latter case, one might expect different results but for predictable reasons. Because of the exploratory mode of the research, it was not possible to determine before the data collection process the most appropriate theoretical base to use to guide project selection. The decision was therefore made to select projects that differed on a range of measures: organisational structure, geographical location (still important in this increasingly wired world), and type of published product. This provided the greatest coverage and best chance of identifying patterns of difference or similarity. The selection technique was based on first investigating candidate projects via a structured literature and Web search.

The research sites were selected to provide a sample of leading-edge projects (details of the projects are discussed below). The geographical range of the selection ended up encompassing both coasts of the United States, the United Kingdom and Europe. No Australian projects (of which there are a number) were included in this phase; the organisation funding this phase (the Victorian Association for Library Automation) required that projects studied be located outside Australia. Together the projects selected provide a good picture of the diversity in the field. Little relevant activity seems to be occurring outside the countries selected. Given the extremely fluid nature of the electronic publishing field, specific comments about the projects are only valid as of early December, 1997 (the end of the data collection process).

Quality control

[Yin, 1998] recommends that researchers continually judge the quality of their case study design. Four tests that are commonly used are to assess if the study has construct validity , internal validity , external validity and reliability . Yin argues that these tests should be applied throughout the case study process: during design, data collection, data analysis and reporting Following these recommendations will "increase the quality of your case study tremendously, and overcome traditional criticisms of the weakness of case study research" [Yin, 1998, p. 242]. Table 8-1 summarises 11 recommended tactics covering theses four tests and also indicates the ways in which the research design and conduct for this case study responded to these recommendations.