Readings in Information Visualization: Using Vision to Think / Edition 1 available in Paperback
This groundbreaking book defines the emerging field of information visualization and offers the first-ever collection of the classic papers of the discipline, with introductions and analytical discussions of each topic and paper. The authors' intention is to present papers that focus on the use of visualization to discover relationships, using interactive graphics to amplify thought. This book is intended for research professionals in academia and industry; new graduate students and professors who want to begin work in this burgeoning field; professionals involved in financial data analysis, statistics, and information design; scientific data managers; and professionals involved in medical, bioinformatics, and other areas.
• Full-color reproduction throughout
• Author power team - an exciting and timely collaboration between the field's pioneering, most-respected names
• The only book on Information Visualization with the depth necessary for use as a text or as a reference for the information professional
• Text includes the classic source papers as well as a collection of cutting edge work
About the Author
Stuart K. Card is a Xerox Research Fellow and manager of the User Interface Research Group at Xerox PARC. He is the author of numerous technical articles and two other books on theories and designs in human-machine interaction. He and his group have contributed to more than 10 Xerox commercial products.
Jock D. Mackinlayis a member of the User Interface Research Group at Xerox PARC, where he has been developing 3D user interfaces for information visualization for over a decade. He received a Ph.D. in computer science from Stanford University and is a member of the editorial board of ACM Transactions on Computer Human Interaction.Ben Shneiderman is a professor in the Department of Computer Science and founding director of the Human-Computer Interaction Laboratory at the University of Maryland. He was elected as a Fellow of the Association for Computing (ACM) in 1997, a Fellow of the American Association for the Advancement of Science (AAAS) in 2001, and a Fellow of the National Academy of Inventors (NAI) in 2015. He is a past recipient of the ACM SIGCHI Lifetime Achievement Award. Dr. Shneiderman is the author and coauthor of many books, technical papers, and textbooks.
Read an Excerpt
Chapter 2: Space6 Equity Trading Analytics
Figure 1 shows a two dimensional image of the equities in the Toronto Stock Exchange index of 35. The Toronto Stock Exchange makes available each equity's order book (i.e. all the current bids and offers and their price and size). For each equity, we are showing the depth of the book. The baseline is the price of the last trade. The offers are above, and the bids are below the baseline. The height above or below is proportional to the price of the bid or offer. The length of a bid/offer's bar is proportional to the size of the bid/offer. This display immediately shows where bids and offers are unbalanced (e.g. more bids than offers) and where there is liquidity.
Figure 2 is a 3D extension of Figure 1. The area of each bid and offer is proportional to the size of the bid and offer. We can use Figures 1 and 2 to do a side-by-side comparison to see what added value a 3D layout provides. The 3D layout in Figure 2 appears to provide faster comprehension, and an accelerated perception. A 3D layout also provides more display area - more elbow room, so to speak. However, 3D is not sufficient. It is not possible to precisely compare two equities, and it is not possible to exactly see the book for an equity located towards the rear of the display. To make 3D work, it must be able to move. To be precise with comparisons, the view of the landscape and elements within the landscape need to move so that the user can see exactly what the values are. The prototypes allow a user to move the landscape and to zoom in to any area of interest.
In the next prototype, we explore how much information can be usefully displayed in asingle screen. The landscape in Color Plate I shows the TSE 300 with the TSE 35 along the right forward edge. The landscape is arranged in an neutral fashion by industry subgroup and alphabetic order within each subgroup. An actual user would order the landscape to correspond to particular interests and models of value. Several different forms of mouse and keyboard driven navigation are supported, including zooming to a point of interest, walking, running and moving to preset fixed points of view.
The TSE 300 landscape can be connected to a live data feed, and can display in real time the liquidity of the most significant portion of the Toronto market. With real time data, the landscape bubbles as trades occur and as the bids and asks are updated. Compared to existing quote screens which can display 20 or 30 equities at one time, this landscape provides an order of magnitude more information.
Of course, liquidity is just one attribute of an equity's performance. Others include net change on the day, volume on the day, volume at a price, trades at the bid, trades at the offer, etc. For each attribute, it is possible to develop a graphical icon or glyph which will visually and precisely communicate the value of the attribute. These graphical elements can be called "signs" because they are designed to display significance. For example, Color Plate 2 shows a trade by trade sign for each equity.
Color Plate 3 shows a bid/offer sign for an equity with numerical data displayed beside the sign. This is an important requirement for information animation applications. Visual perception can be used to quickly see anomalies and patterns. However, at some point, detailed data is needed. The user must be able to point at signs and retrieve the numerical and textual data behind the signs. This capability is called brushing (Cleveland, 1988).
7 Fixed-Income Trading Analytics
The next prototype, in Color Plate 4, shows the Canadian federal bond market. The green yield curve along the left edge shows the Federal benchmark issues as of early January 1992. The first yield curve in red, along the front edge shows the Federal curve as of early February 1992. Using the slider bar, we can play back in a sequence, the values for each day's closing yield curve from January to February. Whatever day is currently displayed is shown in the second yield curve (yellow) along the front edge. At the beginning of the sequence the yellow curve is the same as the green curve. At the end, the yellow is the same as the red. Rotating the scene so that we have a 2D view of the yellow yield curve (Color Plate 5), we can play back each day's yield curve. As we do that, the yellow curve moves and changes over time.
This period of time captures an interesting event - the U.S. Federal Reserve cut interest rates by a full percent. As we playback the data, we see the impact of this event on the Canadian Federal bond market. In the animation, you can see that the curve experiences a large drop, and then recovers to near previous levels.
In the center of the scene, spreads (i.e. differences in yield) are being displayed. The spreads are taken between the start date (green curve) and the current date (yellow curve). Positive spreads are gray, and negative spreads are purple. As we animate the scene, all the spreads move up and down with time.
Several useful conclusions can be made with this prototype. First of all, it becomes readily apparent that the Federal bond market is not an entirely orderly market. Our expectation was that the Federal bond market should be a liquid market with few discontinuities or anomalies. However, as you can see the boundary between positive and negative spreads is ragged. There are positive spreads located among the negative spreads. Further, these anomalies persist for several days at a time.
It is also interesting to note what is happening in this prototype from a human perceptual and cognitive point of view. This is a 20x2O spread matrix displayed over 20 days for a total of 8000 spreads. Compared to a trader's traditional quote screen or spreadsheet display which would show at most perhaps 80 spreads, this is a two order of magnitude improvement in displaying information. The display allows die user, in just a few seconds, to pick out the items of interest. Further, these items are presented in a context which supports informed evaluation. Related 'instrument spreads are shown in the same landscape neighborhood. At any time, the user can click on an anomaly and retrieve numerical and textual data describing the issues and their values.
Another conclusion is that the prototype has the entire field of "spreads" moving. One of the objectives of information animation is to be able to imbed information in motion. The user can see waves move across the surface of the spreads. Subtle differences in the waves, which indicate delays and anticipations in several spread regions, can be perceived and detected quickly.
8 Fixed-Income Risk Management
The next example is a fixed income inventory application. Color Plate 6 shows on a single screen a bond inventory with over 3,000 positions in it. Long positions are in green; short positions are in pink. The left axis shows portfolios and trading groups. The front axis shows time to maturity. Height is used to show the value of the positions. Along the front of the landscape is a total line that totals across trading groups. At the rear is a yield curve based on the bid yield of a set of benchmarks.
Color Plate 6 is displaying a profit/loss for each position for a yield curve shift scenario. Other models of asset value can also be displayed (e.g. weighted price value of a basis point or benchmark equivalent).
The user can point at one of the positions and retrieve fully detailed descriptive numerical and textual details related to that position's size, issue and issuer.
Animation is used in this landscape to help assess market risk (i.e. risk due to change in interest rates). The yield curve can be moved, in even or uneven shifts, and as it moves, the impact on the inventory's projected profit and loss can be assessed. Projected P+L values change by increasing or decreasing in size. Users can quickly see where the inventory is hedged (i.e. insensitive to changes in rates) or where it is exposed and by what degree.
Projected P+L is calculated using pre-computed fixed income analytical parameters such as the dollar value of an 01 (i.e. unit value for a change of 1/100 of a percent in interest rates) and convexity (i.e. second order approximation for sensitivity due to change in interest rates). The scenario P+L calculations are done rapidly so that the P+L value changes interactively with a change in the yield curve.
There are several ways a user may interact with this landscape. One or several bond issuers can be selected from a list of all issuers. The corresponding positions in these issues are then highlighted within the inventory so that the user can see what is held and where it is held. The total line then shows the total for the selected set of issuers. The market risk scenarios can be performed on the selected set.
Another type of query example is a filter based on size of position and implemented using a slider bar. The user can filter out all small long and short positions, so that landscape displays only the large magnitude positions.
This risk viewing landscape provides several conclusions. An on-line system could display perhaps 20 to 40 positions per screen. A 3D landscape displays 3,000 or more positions per screen. Using query and filtering, it is possible to highlight patterns that may be hidden in a numerical display. We believe a 3D visual approach provides more insight in minutes than traditional computer numerical displays could provide in hours.
9 Additional Examples
Information animation applications can provide significant value in many areas of the securities industry, The risk viewing application can be expanded to instruments and markets beyond fixedincome, including equities, derivatives, and foreign exchange. In each case, the common elements in the application are positions, models of asset value, parameterized models of risk sensitivity, and scenario projections. Risk viewing starts with being able to quickly see thousands of positions on a single screen. Effective risk visualization requires direct manipulation of such risk parameters as interest rate risk, volatility risk, currency risk, and credit risk. These visual risk models more easily allow control and guidance of risk exposure over a wide variety of scenarios and stress tests. Simple combinations of changes to risk parameters will quickly reveal exposed positions and help suggest more effective risk management strategies.
Another information animation application can be developed for trading in the OTC (over-the-counter) market. NASDAQ level 2 market-maker data provides bid and ask information for all market makers in an equity. Generally, there are 10 to 50 market makers per equity. Numerical equity trading displays now in use are limited to showing marketmaker activity for just one equity at a time. Further, the user is able to see only 15- to 20 market makers at once, and must page back and forth to see other market makers. Considering that an individual trader...
Table of Contents
1. Information Visualization
4. Focus + Context
5. Data Mapping: Text
6. Higher-Level Visualization
7. Using Vision to Think
8. Applications and Innovations
9. Conclusion Bibliography Index