Using Space Effectively: 2D
From Visualization Sp06
Lecture on Feb 7, 2006
Readings
- Hyperdimensional data analysis using parallel coordinates, Wegman (jstor)
- Generalized fisheye views, Furnas (acm)
- Pad++: A zooming graphical interface for exploring alternate interface physics, Bederson & Hollan (acm)
- A framework for unifying presentation space, Carpendale & Montagnese (acm)
- Chapter 11: The Cartogram: Value-by-Area Mapping. In Cartography: Thematic Map Design. Dent (handout)
Optional Readings
Contents |
Lesliei - Feb 06, 2006 05:36:02 pm
I thought the parallel coordinates reading was really interesting and dug up this applet that implements it. It took some work to figure out the interface, but here's what I found. The name of field label should actually go on the pulldown menu below the textbox. You can choose different fieldnames from this menu, and the red dot at the bottom of the visualization tells you which vertical axis corresponds to the field name you chose. The blue dot at the top controls how the lines that correspond to the data (I'll call them data-lines for short) are colored. E.g., if you move the blue dot to the third vertical axis, a data-line will be colored black if it intersects the third axis at the top, red if it intersects the axis at the bottom, and a combination if it falls in between. To get the value of a data-line at a vertical axis, move the mouse over the data-line / axis intersection until the data-line turns cyan. Finally, you can zoom vertically by entering values in the corresponding widgets or by clicking and dragging a vertical range on the visualization.
Bryan - Feb 07, 2006 09:49:42 am
I also liked the idea of parallel coordinates. I will admit that it is much harder for me to tell the difference between a correlation coefficient of 0.8 and 0.2 than in a standard, 2D scatter, but I honestly can't think of another way of showing it on a static 2D page. My biggest problem with the approach though is that I don't think it should be static. Statically, there is an ordering to the dimensions that doesn't exist in a standard scatter, making it hard to compare two dimensions that are not adjacent. It would probably work better as an interactive visualization on a computer, ideally one that allows reordering of axes by dragging and selection of any two of those axes to view as a scatter plot.
Pushkar - Feb 07, 2006 01:42:47 pm
In Carpendale and Montagnese's "A Framework for Unifying Presentation Space", the authors argue that the concept of 'folding' (see bottom of pages 4 and 7) provides a visual continuity between the magnified (and/or distorted) region and the context (or unmagnified/undistorted) region.
Is this visual continuity really necessary? It looks good for zooming into actual geographic maps of real-world data, but does the distorted info. on the 'walls' of the folded region (see fig. 15) really convey useful information?
I feel like the visualization may be adding unnecessary information in the case of non-map 2D data. Besides, it's geometrically inaccurate in case of the Manhattan lens - we're trying to map an infinitesimally small region of space (the region boundary) onto a finite 2D surface.
Nchentan - Feb 09, 2006 01:50:17 am
I have some comments for the Cartography: Thematic map design article. Personally, I have never seen this type of diagram before and found it to be quite visually interesting. However, it seems to me that it might be rather bad at showing quantitative information. As we learned in class, perception of area is not linear even when the objects have the same shape. In this case, the shapes are different, which seems to make area estimation even less accurate. I think that it's might be a good visualization to use only when we want to show that a quantity is high in regions that are geographically small or when we just want to show the ordering.
AaronHoover - Feb 11, 2006 11:01:19 am
I had the exact same response to the cartogram as Nchentan. I too hadn't seen cartograms before, and I even had a little difficulty interpreting the ones presented in the reading. I was wondering about the effectiveness of the cartogram given the perceptual difficulties of judging quantitative information from areas. However, I did find the cartogram depicting the presidential election votes sort of generally useful for comparison. But, I think its effectiveness may have been largely due to context and knowing beforehand that it represented a distortion of a map of the U.S. It seems to me that the usefulness of cartograms probably depends heavily on the reader's prior experience with them.
Brien - Feb 14, 2006 11:48:34 am
I think Pad++ is rocking. The paper above and others on their site -- especially the ones describing derived apps geared towards children -- all note net preference for zuis over traditional apps. Even though supposedly Sony licensed this toolkit, I wonder why no major applications have come out using similar concepts. For example, I've never seen a zooming web browser, even though I think they would very useful in a library. It can't be technical reasons ... Furnas has another paper, "Space Scale Diagrams", that shows an interesting way to create a 1D fisheye using a projection from "1D+1D". I like how his paper above demonstrates the applicability of fisheye to non-graphical apps, like a code editor. It seems the issue of losing context might be much worse for code than graphics, though.
Raymond - Feb 14, 2006 12:10:44 pm
I thought that the nomograms (esp. the one used for sailing) were really cool. They were the first time I've seen them before, yet they were easy enough to understand to provide complex results without understand the calculations behind it. I'm also amazed at intricacy and complexity of making a nomogram, as I'm sure that even you had all the data, it would be hard to create a visually effectice one. Unfortunately, I'm not sure how useful they would be now, as computers and java applets would probably be used instead of nomograms or slide rulers.
Yi-Tao - Feb 21, 2006 04:09:38 pm
AaronHoover - I'm surprised that you didn't have trouble interpreting the cartogram, because I didn't realize it was a the US when I first saw it. I find that cartograms are harder to interpret if the underlying object already has a shape (i.e., a map of the states). For the most part, the cartograms are useful for comparisons, but I do agree that some of the polygons are insanely difficult to compare.
Brien - I think the main reason that zooming applications are not widespread is because we tend to use a hierarchical approach (think nested folders). We don't write subheadings in ever decreasing size; so even if zooming might be a better way of doing it, it wasn't implemented first.
