Analyzing Case Study Data - Yin

The importance of having a clear analytical strategy is highlighted in this chapter of Yin (2009).  It is the least developed and most difficult aspect of conducting a case study and 'much relies on the investigator's own style of rigorous empirical thinking, along with the sufficient presentation of evidence and careful consideration of alternative interpretations' (p.127).
There is a short summary of the use of computer assisted tools but a reminder comes that they are only tools and only assistants, how the investigator manipulates the data is more important. It is recommended to first 'play' with the data and some ways in which to do this are summarised. For examples to create arrays; make a matrix of categories; use flowcharts and graphics; frequency of events; statistically analyze events; organising data chronologically.

All data has a story to tell and the needed analytical strategy will help you craft the story. Yin describes four general strategies.

Relying on theoretical propositions. Those that led to your case study, shaped the research questions and the literature review and therefore led to new hypotheses or propositions.

Developing a case description. Whilst less preferable to the previous strategy, a descriptive framework in which to organise the data can sometimes be useful particularly when the data has been collected without any theoretical propositions being made. Such a framework should have been developed before designing the data collection instruments. A descriptive approach may help to identify any causal links which can then be analysed.

Using both qualitative and quantitative data. There are two reasons why quantitative data may be relevant to your study. a) the data may cover the behaviour or events that your study is trying to explain. b) it may be related to an embedded unit of analysis within the broader study. Using statistical techniques to analyse this quantitative data at the same time as analysing the qualitative data will strengthen your study.

Examining rival explanations. This can be done within all of the previous three strategies and will strengthen the analysis if collection of evidence regarding 'other influences' is carried out. These rivals can be categorised as craft rivals (including null hypothesis, researcher bias and threats to validity) or real-life rivals. The more rivals a study addresses and rejects, the more confidence can be placed on findings in a case study.

The next section outlines five analytic techniques. Each one  needs to be practiced and the case study researcher will develop their own repertoire over time to produce compelling case studies.

Pattern Matching. This can strengthen a study's internal validity by comparing empirically based patterns with predicted ones. The following pattern types can be used:

• nonequivalent dependant variables as a pattern
• rival explanations s patterns
• simpler patterns

No matter the pattern type chosen, the more precise measures you can obtain, the stronger your argument/case study will be.

Explanation building.  When the data is analysed to explain the case. A parallel method for exploratory case studies can be used if the aim is to develop ideas for further study rather than in this case, to make conclusions. To explain something we stipulate a presumed set of causal links as to why or how it happened. Due to the imprecise nature of narratives, 'the better case studies are ones which have reflected some theoretically significant propositions' (p.141). The eventual explanation is likely to be a result of a series of iterations whereby the evidence is examined, theoretical propositions revised and the evidence examined again from a new perspective. As in pattern matching the aim is to show how rival explanations cannot be supported. The author notes however that this approach is 'fraught with dangers' and refers back to earlier advice to strengthen the study such as following a case study protocol, creating a case database and the following of a chain of evidence. If constant referral to the original source of inquiry is made and to the possible alternative explanations, the researcher is less likely to stray along a divergent path, during iterative cycles.

Time-Series Analysis. If there is just one dependent or independent variable we call it a simple time series. A mixed pattern across time can give rise to a complex time series and tracing events over time is known as a chronology. Care must be taken with this type of analysis not to simply describe or observe trends over time (this would be known as a chronicle) but rather to look for causal inferences.

Logic Models. This technique involves matching empirically observed events to theoretically predicted events and is differentiated from pattern matching because of their sequential stages. There are four types, related to the unit of analysis chosen.

• Individual-level 
• Firm or organisational-level (linear sequence)
• Firm or organisational-level (dynamic, multi-directional sequence)
• program-level

In each type, you should define your logic model then 'test' the model by analysing how well your data supports it. (p156).

Cross-Case Synthesis. This techniques treats each case as a separate study. One example is the use of word tables to display data for each case according to some framework. Then cross-case conclusions can be made. However care must be taken in developing 'strong argumentative interpretation' (p160).