Patton has a good discussion of extreme cases as a sampling technique in his book, Qualitative Research & Evaluation Methods. But as for analysis, I am not aware of much that has been written on that topic.

Can you say more about how you are combining your qualitative and quantitative data, and the role that extreme cases play in your overall project?

Article The Case for Selecting Cases That Are Deviant or Extreme on ...

Article Case Studies: Types, Designs, and Logics of Inference

Article Deviant Case Analysis in Social Science

Article Dependency Theory and Taiwan: Analysis of a Deviant Case

Article The Paradoxical Value of Deviant Cases: Toward a Gendered Th...

David L Morgan thanks for your advice, I'll search for that author and the work.

I'll try to answer your question about the methodolgy - it has been a bit of journey.

It is a very explorative study where I interviewed a variety of engineers and asked them to provide examples of creative problem solving in their roles, and also their thoughts on creativity in the engineering space. All participants completed 5x psychometric measures and provided d a large amount of demographic data. Myself and an interrater analysed the qualitative data but we were also able to quantify certain aspects on a ranking scale (ordinal) and I then performed a statistical test of trend (n=25, so good qualitatively, but weak statistically). Fortunately, the psychometric measures I used all have norms/benchmarks so could also be categorised as ordinal. Once that was done and some interesting patterns (and some significance) were found, conversations with peers and supers led to the decision it was the extreme cases that were actually of interest - so I'm going back to the cases and all the information collected to try find out what is going on between those who score extremely 'positive' and those at the other end.

Methodologically - it sounds messy as I type it out here, and I'm obviously still learning - but it has been rigorous and novel, and I hope that the pattern I think I have found is actually there, which would be of use to the engineering education and workplace sectors - definitely a discussion piece if anything. This is part of my doctorate so a good opportunity to learn some new skills and test out some methods. Please share with me if you have any thoughts/advice?

- Michelle

Have you looked at a scatter plot of your two types of responses? That would show you whether there were extreme cases that were "detached" from the rest of the distribution.

David L Morgan Yes, certainly did - that is how I identified the interesting pattern. It was hard with a 3x3 frequency table but the sample size was small enough to identify the cases easily enough.

Yes, if you only have a limited number of categories, then a table would provide a useful summary. My personal opinion is that the most interesting cells in such a table would be either of the two off-diagonal corners. Those would contain a combination of very low on one variable and very high on the other.

David L Morgan - thank you - I must be on the right track as everything you have asked I have done!

David L Morgan Cristian Ramos-Vera thank you for your help all this time ago. I will be able to share my publication soon - I got there in the end.