Abstract:
Two-dimensional graphical representations of data have been the de-facto standard for conveying information in statistics, whether the data of interest are univariate, bivariate or inherently multivariate. However, little exploration has been made into the realm of the third dimension. This may be because of the results of the Cleveland-McGill perceptual experiments or to hardware limitations that have severely hindered its use. Even less has been done to address the question of how to enrich the level of interaction with three-dimensional plots. This thesis represents a first step to fill these gaps. First, a prototype was built to help identify the essential elements for building a three-dimensional graphics application framework. The knowledge and the lessons learned will then pave the way to a future software product that will allow statisticians to build highly interactive graphical applications. The prototype itself will also, as a very important corollary, facilitate the building of perceptual experiments, effectively extending Cleveland’s work by introducing various visual cues and scene navigation methods. In the second half of the thesis, the Model-View-Presenter (MVC) family of GUI (Graphical User Interface) architectures are studied in some depth. One of its later incarnations, Taligent Model-View-Presenter, was found to be superior to the others, and is therefore a likely candidate for the next generation data visualisation framework. The adoption will ensure a more consistent way of building both visualisation and GUI components, and improve maintainability and re-usability at the same time. In particular, Taligent MVP readily provides multiple undo/redo for any visualisation applications that are built by it. The feature is rarely seen on other statistical graphics software and is worth further exploration.