Abstract:
|
Results are most useful when we can trust, understand, reason about, reuse, and extend them. In the case of static (script-generated) results, provided that the data and software used are available, the script which generated a result embodies both the ability to computationally verify that result - provided the data and software are available - and the list of the methods applied and computational steps used to create it. This allows the reproducibility of a static result to provide, in principle, assurances about our ability to trust, understand and reason about the result. When discussing reproducibility of a particular state in an interactive graphic, however, the analogs to these two aspects of the generating script are decoupled; they are embodied by the state itself, and the history of the user's actions, respectively. We will present work on extending a reproducibility and discoverability-based result management platform designed for static results and applying the underlying conceptual framework to states within shiny-based interactive graphics in R.
|