University of Maryland Department of Computer Science Spring 2007 CMSC 734 Information Visualization Prof. Ben Shneiderman
Midterm exam - open book (papers), open notes, no discussion, no laptops
1) The book Illuminating the Path describes an iterative analytic process with the sense-making loop with four phases: gather info, produce result, develop insight, re-represent. Consider an information visualization example (web site or application) and critique it according to how well it supports these four phases, especially how well it supports discovery (develop insights) and how well it supports presentation of results to others (re-represent). Cite specific positive and negative features (10 points).
2) MacEachren et al. make a strong point of using entropy as opposed to correlation coefficient. Draw example 2-D distributions that show (15 points) 2a) low entropy and low correlation 2b) low entropy and high negative correlation 2c) Is there a reason why high entropy is incompatible with low correlation?
3) The rank by feature framework was designed to deal with 1D and 2D distributions. But now consider the design of a rank-by-feature (RBF) interface for a temporal data set, like the 8 years of George Bush’s State of the Union speeches (SOU8) (for January 2008 NY Times). It should allow users to type in any word and see the distribution of usage over 8 years, and then it should offer the RBF interface to assist discovery of interesting words. 3a) Give five criteria for ranking the distributions generated by every word, with an explanation of why each is interesting (10 points) 3b) for one of these, describe the algorithm or formula for the ranking. (10 points) 3c) In advance of making the SOU8 website public, the designers would like to conduct an evaluation by hiring a usability company. Write the detailed request for proposals for this evaluation, describing the goals, specifying the design for a two-week process, and the format+content of the final report (20 points).
4) Design a visualization to show the patterns of grades in a course with 20-200 students, so as to help the professor assign final grades [A (>=90), B (>=80), C (>=70), F (<70)]. There is an early project (30 percent), a midterm exam (15 percent), a semester project (40 percent), and final exam (15 percent). All grades are in the range of 1 to 100, but you should acount for the situation that a student may miss one of these grades. The professor would like to do more than just use a spreadsheet with weighted average of grades. 4a) draw your screen design including labels and your legend (15 points) 4b) describe two interactive features may help the professor make more fair final grades (Hints: might reduce impact of one low grade, be generous to students whose grades increase, or make step up in letter grades where there is a gap between averages) (10 points)
5) A dataset contains information on 11 UMd Colleges (CMPS, ENGR, CLIS, BMGT, ARHU, etc.), which contain 200+ Departments, and 2000+ faculty. For each faculty member there is a row of data that includes name, rank (Asst, Assoc, etc.), years since Phd, and current year’s annual salary, number of refereed papers published, number of grad students supervised, and grant support income. Consider 2 questions that might be interesting to explore from this data for the provost, deans, or department chairs. Write the 2 questions, indicate the person, and then sketch the treemap you would suggest including legend for each question (10 points).