Single complex glyphs versus multiple simple glyphs

Abstract

Designers of information visualization systems have the choice to present information in a single integrated view or in multiple views. In practice, there is a continuum between the two strategies and designers must decide how much of each strategy to apply. Although high-level design guidelines (heuristics) are available, there are few low-level perceptual design guidelines for making this decision. We performed a controlled experiment with one, two, and four views to evaluate the strengths and weaknesses of these strategies on target detection and trend finding tasks in the context of multidimensional glyphs overlaid onto geographic maps. Results from the target detection tasks suggest that visual encoding is a more important factor when detecting a single attribute than the number of views. Additionally, for detecting two attributes, the trend indicates that reusing the most perceptually salient visual feature in multiple views provides faster performance than an integrated view that must map one of the attributes to a less salient feature.

A Systematic Review of Experimental Studies on Data Glyphs

“…Note that papers can fall in multiple cells. Since PlanningLines [35], weather vanes [36], [37], shapes [38], [39], roses [40], themes [41], arrows [42], Motifs [43], [44], flowers [45], and MILSTD2525 glyphs [46] use a unique encoding and are not compared to other glyphs, we positioned them slightly apart in the "One-to-One Mapping" category.…”

“…Patients rated by psychiatrists [21] Minnesota Multiphasic Personality Inventory Medical data [32] unknown Medical images [56] unknown Cars dataset [34] http://davis.wpi.edu/ xmdv/datasets/cars.html Project plans [35] unknown MM5 weather information [36] unknown Weather information [37] NCEP forecast model Weather information [42] Operational Regional Atmospheric Prediction System Modified U.S. census data [38] unknown Financial data [40] Investment in education USA (2008) Financial data [84] Wall Street Journal Index (1974 and 1975) Financial data [85] Standard and Poor's firm list (1974 and 1975) Classical music data [41] unknown k data [43] Lostpedia wiki edits Network data [44] 10 best ranked movies (IMDb) Google search results [45] http://www.google.de Marathon runners [50] unknown Power plant statistics [53] unknown Audio information [62] One laptop per child sound library Biological data [70] unknown Economic variables [77] U.S. Department of commerce & labor Tensor data [92] DTI dataset possible confounding factors, so as to better reason about changes in performance when modifying specific aspects of the same design [91], or when comparing it with other representations [48], [49], [50], [54], [64], or layouts [55].…”

“…This is perhaps not surprising as glyphs are often meant to provide quick overviews over a large number of varying multi-dimensional data points-and the use of synoptic tasks may reflect the authors' desire to test glyphs in a realistic use context. Multi-dimensional [21], [34], [36], [37], [38], [40], [42], [45], [46], [48], [49], [50], [51], [53], [60], [65], [72], [79], [88], [90], [91] [32] [38], [84], [85] [17], [33], [45], [47], [52], [57], [58], [59], [61], [63], [64], [66], [73], [74], [75], [76], [77], [78], [80], [86], [87], [89], [92] [37], [39], [40], …”

“…Study missing mappings: In Table 3 we refer to 50 of the 64 studies examined, having left out the two one-to-many mappings [32], [33] and the twelve that were not compared to other designs [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46]. Looking at the table there is still clearly an imbalance in what kind of data encodings have been comparatively tested.…”

On Open Problems in Biological Network Visualization

“…When multiple views of the same data are used simultaneously, brushing is typically associated with linking, i.e., brushing of elements in one view affects the same elements in other views. First results that try to compare both approaches [25], i.e., multiple views and attribute integration, cannot be directly used in our complex case of biochemical networks.…”