I suggest both! I think it makes sense to have a dataset we work on with the full group, which we move through at their pace (guided perhaps by our own, private roadmap of how we imagine the process). However we won't have time to cover much this way, as Mike says. It would be great to be able to take some time and walk through a few "case studies", perhaps to illustrate or reinforce concepts as we try to communicate them. (Quotes around "case studies" only because I imagine a treatment that is less close and careful than Mike's excellent case studies. ) I have numerous examples from my own work with my colleagues from which to choose, we could talk about it.
Additionally, it would be VERY interesting to leave time for participants to discuss their own projects. This could be done in small groups, perhaps towards the end, and would certainly yield up some interesting examples and questions
@aamd suggested a mix: