Selection sequences - Interactive analysis of massive data sets

One of the main tasks in analyzing massive data sets is to condition views, i.e. graphs and/or statistics, of the data on subsets. One recent approach to do so are Trellis Displays, introduced by (Becker et al. 1994). But whereas Trellis Displays are static and show all views of all subgroups at a time, the exploration of data sets is iterative and it often focuses on particular subgroups. Selection Sequences are an intuitive extension to the established linked-highlighting paradigm, which e.g. is implemented in DataDesk or SAS-Insight. Selection Sequences store the whole hierarchical path of a selection and allow an easy editing, redefinition and interrogation of each selection in that path. This enables the user to analyze even massive data sets without losing the aim he was heading for. Since all selection parameters are stored in a sequence, a generalization to graphical data base queries is easy to achieve. A first implementation of Selection Sequences (cf. Hofmann et al. (1997)) can be found in the research software MANET.