Visual Analysis of Multi-Parameter Distributions across Ensembles (Englisch)

For an ensemble of data points in a multi-parameter space, we present a visual analytics technique to select a representative distribution of parameter values, and analyse how representative this distribution is in all ensemble members. A multi-parameter cluster in a representative ensemble member is visualized via a parallel coordinates plot, to provide initial distributions and let domain experts interactively select relevant parameters and value ranges. Since unions of value ranges select hyper-cubes in parameter space, data points in these unions are not necessarily contained in the cluster. By using a multi-parameter kD-tree to further refine the selected parameter ranges, in combination with a covariance analysis of refined sets of data points, a tight partition in multi-parameter space with reduced number of falsely selected points is obtained. To assess the representativeness of the selected multi-parameter distribution across the ensemble, a linked side-by-side view of per-member violin plots is provided. We propose modifications of violin plots to show multi-parameter distributions simultaneously, and investigate the visual design that effectively conveys (dis-)similarities in multi-parameter distributions. In a linked spatial view, users can analyse and compare the spatial distribution of selected points in different ensemble members via interval-based isosurface raycasting. In two real-world application cases we show how our approach is used to analyse the multi-parameter distributions across an ensemble of 3D fields.