SIMS 247 Assignment 3

Issued: Feb 24
Due: March 3
Format to turn in: hardcopy or pointer to a url.

Last modified 2/23/98

Update on timing of assignments

Assignment 4 will be issued on March 3 and due on March 17. It will involve using Pad++, so start reading the paper by Meyer, the two papers by Bedersen, and the Pad++ manual.

The requirements for the midterm project will be distributed March 17 and the results due April 7. The requirements for the final project will be distributed April 7. Presentations on the final projects will be due the week of May 5. Project reports will be due May 14.

There will probably be another short homework assignment during the time that you are working on the final projects.

Questions

(1) Magic Lenses for Querying on Data Graphics

The original formulation of the Toolglass/Magic Lenses idea (Bier et al.) demonstrated how it could be applied to dramatically change the interaction within a drawing program. The later formulation (Fishkin & Stone 95) showed how the idea of movable, transparent, combinable filters could be used to enhance interaction with data graphics like scatterplots.

The following questions refer to the Fishkin & Stone 95 paper. This is in the reader but is also accessible on the web if you find the figures hard to see. http://www.acm.org:82/sigs/sigchi/chi95/Electronic/documnts/papers/kpf_bdy.htm

(a) Consider operations on scatterplot data such as showing all counties with a median income larger than $30,000. Describe how this is done with magic lens filters vs. with the interaction facilities available in Spotfire.

(b) The call-out filter (color plate 2) is an alternative to which SpotFire operation(s)? Which of these do you think is more useful, and why? (Or if one is more useful in some situations than others, state what the different situations are.)

(c) How can the magic lens filters distinguish between nominal and quantitative information? (Hint: see figures 4a and 4b.)

(d) In class we talked about how we might create a Show Outlier magic lens filter. Describe what such a filter would do and illustrate it with an example that of the census data we worked with in Assignment 2 (you can hand draw this if you like, and you can use fake data if you don't have any examples of outliers in that data handy).

(e) Describe how the filter of (d) might be overlaid with some other lens to act as a Show Everything Except Outliers filter.

(2) Focus + Context and the Table Lens

(a) What are the roles of focus and context in the Table Lens (Rao & Card 94)? What happens when something is taken out of focus?

We've seen at least two ways to look at the baseball statistics information: brushing and linking across scatterplots, bar graphs, and histograms (in the EDV system, and Spotfire to a lesser extent) and via sorting columns of information in the TableLens (Rao & Card 94). (I'll give a handout containing the color plates missing from the reader; EDV is discussed in lecture notes and can be viewed at http://www.bell-labs.com/user/gwills/EDVguide/bb.html .)

(b) What kind of interaction operation is possible in the Table Lens that is not done in standard brushing and linking as seen in EDV and SpotFire? (I am not referring to the fact that one uses gestures, but rather in the kind of operation the gestures are used for.)

(c) How are outliers, say in salary vs. number of hits, found in TableLens vs. EDV? How are patterns using two or more attributes (such as seen for put-outs vs. assists) discovered using the TableLens vs. EDV? (Hint: this is basically the same as the answer to (b)).

(d) Can more information be seen at once in the Table Lens, in EDV, or are they equivalent?

(3) Graphing Networks

Consider the following ways of graphically displaying trees and networks (remember, a tree is a special case of a network, and graph (confusingly) is another name for network).
• Scrolled expandable/contractable hierarchical displays (like those seen in the Mac and NT file manager browsers)
• Hyperbolic browser (Lamping et al.)
• Cluster graphs (Eades & Feng)
• Force-directed placement (Amir, Hendley et al.)
• Multitrees (Furnas & Zack)

Consider the following information sets with network structure:

• Directory and file structure on a computer system
• The Yahoo category network
• A pattern of web page accesses across time (as seen in the Tauscher & Greenberg paper of assignment 1)

For each of the network display types, discuss how useful they may or may not be for each of the information sets.

(Bear in mind that Yahoo is not a strict hierarchy; those categories marked with an @ sign have a parent link pointing to them from somewhere else. However, you may treat it as a hierarchy if you like (hint: consider multitrees here).)