next up previous contents index
Next: 10.3 Concepts of Interactive Up: 10. Interactive and Dynamic Previous: 10.1 Introduction


10.2 Early Developments and Software

There is a strong history of statistical graphics research on developing tools for visualizing relationships between many variables. Much of this work is documented in videos available from the ASA Statistical Graphics Section Video Lending Library at http://www.bell-labs.com/topic/societies/asagraphics/library/index.html. Currently, these historical videos are upgraded into a digital format and will be available on CD or DVD later in 2004.

Additional material on statistical graphics can also be found in journals such as ''Journal of Computational and Graphical Statistics'', ''Computational Statistics'', and ''Computational Statistics & Data Analysis'' and in ''Computing Science and Statistics'', the proceedings of the Interface conferences. The following paragraphs only serve as a basic overview for readers unfamiliar with dynamic statistical graphics. They are not intended as a full introduction into this topic.

A video clip of the successive stages in a multidimensional scaling algorithm ([111]) is one of the first examples how to apply dynamic statistical graphics. A second example by [47] shows an interactive search for a structured two-dimensional projection in five dimensions where three of the five dimensions are noise. PRIM-9 (Picturing, Rotation, Isolation and Masking in up to $ 9$ dimensions), documented in [72] and [73], is the landmark example of early dynamic statistical graphics. Projections formed the fundamental part of the visualization system and were complemented with isolation and masking. A good explanation of the importance of projection as a tool for visualizing structure in high-dimensional data can be found in [80].

One major breakthrough in using projections for visualizing higher dimensions was made by [7] in his work on the grand tour. The grand tour, further exploited in [20], in an abstract sense shows a viewer all possible projections in a continuous stream (which could be considered to be moving planes through $ p$-dimensional space). Several possibilities for ''showing all possible projections'' were explored in the original work, but the most successful method to arise from it is based on interpolating between random planes. Another common approach to displaying high-dimensional data can be found in [13] where data is plotted in a scatterplot matrix, i.e., a matrix of pairwise scatterplots. Users can do linked brushing among the plots, i.e., mark points with different symbols and colors, while this information is also immediately displayed in all related (linked) plots.

The historical development of interactive and dynamic statistical graphics is well documented in a series of books and articles. [46] and [68] can be placed somewhere inbetween Tukey's original idea of EDA and the beginning of modern dynamic and interactive statistical graphics. [185] is a collection of papers presented at a workshop sponsored by the Office of Naval Research (ONR), held in Luray, Virginia, from 24 through 27 May, 1983. About half of the papers are related to statistical image processing while the other half is related to (interactive) statistical graphics. [49] contains a collection of papers about dynamic graphics for statistics, originally published between 1969 through 1988. This book is a very good reference to see the progress in dynamic graphics concepts and software over two decades, starting from the very early stages through the late 1980's. [26] is based on the proceedings of the Institute for Mathematics and its Applications (IMA) 1989 summer program on ''Robustness, Diagnostics, Computing and Graphics in Statistics''. An earlier ''Handbook of Statistics, Volume 9, Computational Statistics'', edited by [130], contains several then state-of-the-art overviews on interactive and dynamic statistical graphics, most notably the chapters by [184] and [201]. [123] dedicate two (out of six) chapters of their book to dynamic graphics - one being an overview and one discussing applications. [162] is fully dedicated to the theory and applications of interactive statistical graphics. [193] contains reviews of software for interactive statistical graphics.

Major statistical journals often dedicate special issues to interactive and dynamic graphics, e.g., ''Computational Statistics'' (Volume 14, Issue 1, 1999) on ''Interactive Graphical Data Analysis'' and ''Computational Statistics & Data Analysis'' (Volume 43, Number 4, 2003) on ''Data Visualization''.


next up previous contents index
Next: 10.3 Concepts of Interactive Up: 10. Interactive and Dynamic Previous: 10.1 Introduction