1999 Internet Usage Survey explored in Table Lens and Spotfire
Erica Kolatch
CMSC 838B: Information Visualization Application Project
February 28, 2001
Introduction
The Internet is quickly developing into a common method of information gathering, communication, and entertainment. However, questions abound on the effect of the Internet on society. There are also significant questions concerning gender and race differences when discussing the use of information technology in general and the Internet in particular. Several studies have been completed or are underway to study how technology and the Internet are being used.
Data
WebUse (UMCP), a scientific center for the study of the Internet offers several datasets on Internet Usage. One of these is the UMD Internet Usage Survey conducted in 1999. Information was gathered on Internet or e-mail and home computer use, “yesterday”. A yes answer led to further questions. In addition, general demographic data including age, race, and income were collected. Slightly over 1000 respondents were surveyed. To ease data use I first translated it from numerical coding to string data, and for analysis purposes, I also created two composite columns. The first, Total Internet and E-Mail Usage, sums questions Q1E-Q1H, which relate to minutes spent using Internet and E-Mail. The second, Total Home Computer Usage, sums questions Q2B-Q2H, which relate to minutes spent using the home computer.
Methodology and Tools
My initial plan was to use a tool under development in the HCIL at the University of Maryland, called Dynamaps. Dynamaps is designed to aid in manipulating map-related data. [1] Since the data set covered 48 states, I thought this might show and confirm known patterns about technology and Internet usage. However, Dynamaps is a work in progress, and as such was unavailable for this application. I therefore turned to two other visualization tools. The first was Table Lens, a tool originally developed at Xerox Palo Alto. [2] Table Lens is a visualization tool for large datasets, accommodating 30 times as many cells as can be seen at one time on a normal screen. The second tool was Spotfire. [3] This tool provides multiple methods for visualizing data and exploring the relationships between attributes.
Observations – Table Lens
One of the valuable features of table lens is the ability to fix multiple variables in a sorted position, and then vary the sort on other variables. Figure 1 shows an example of this phenomenon. The data was first sorted by sex, and then by whether the respondents used the Internet or e-mail on the previous day. Finally, it was sorted by the amount of time spent using e-mail at work.
In figure 1, the first two dark gray columns are fixed, and the next five are in focus. Eyeballing the results suggests that although more women than men responded to the survey, an equal number of men and women reported using the Internet or e-mail, and men and women spend approximately the same amount of time using e-mail at work. Similar results appear when looking at the amount of time spent using the Internet for work.
Figure 1
As an alternative to choosing gender as one of the primary attributes, figure 2 uses technology. Here the columns “Used Internet or e-mail yesterday” and “Used Home Computer Yesterday were selected. Figure 2 examines how gender and income relate to the use of technology. The yellow bars are Male, the red Female. For technology users the income ranges are fairly similar, but there are more low-income non-users of technology, and there is more evidence of a gender gap.
Figure 2
Finally, in order to show some of the other features available on Table Lens figure 3 is sorted by race, and to show the power of the focal lens, it focuses in on non-white users of technology. The answers for these respondents can be seen clearly, while those of the other respondents remain in a general grouping. For comparison purposes, figure 4 shows the same sorting criteria without the focal cells.
Figure 3
Figure 4
Observations – Spotfire
In contrast to Table Lens, Spotfire looks at points of data based on the values of specific attributes. In figure 5, we can see confirmation that technology usage is related to education. As education levels rise, technology usage rises. Technology usage is on the x-axis, education on the y-axis.
Figure 5
Similarly, figure 6 shows that technology usage among respondents was highest for ages 20 to 50, with an almost equal split between men and women. Usage is on the x-axis and age is on the y-axis.
Figure 7
Spotfire could also be used to make comparisons over multiple variables (or dimensions) using either 3d visualization or parallel coordinates. For sample screen shots please see the Appendix.
Lessons Learned / Critique of Tools
Both tools allowed for easy manipulation of the data. However, Table Lens’ strength was in fixing the sorted order of one or more attributes and then seeing the effect on other attributes. This technique works well with scaled data. Multiple yes/no questions make for confused results. Spotfire inserted all selected data points relating to a particular set of attributes, and general trends were visible, but it was more difficult to identify smaller sets while still visualizing the whole scale of the data set. In particular, when trying to examine usage based on race, since white was the most common response it tended to hide other choices. However, if white is filtered out, the result gives a skewed view of the response set.
These tools are extremely useful for portraying visually what has been developed through statistical methods. They are also useful for spotting trends that might not otherwise be seen, or confirming visually information gleaned from other sources. Both Table Lens and Spotfire are successful in showing this for the Internet Usage data. However, I still believe that it would be interesting to see a state-by-state visualization of Internet usage such as that provided by Dynamaps. Regional digital divide issues would become more visible, something impossible to do in these visualizations.
[1] "User Interfaces for the U.S. Bureau of Census: Online Survey Interfaces and Data Visualization" http://www.cs.umd.edu/projects/hcil/census/ Back
[2] Rao, R., and Card, S.K. "The Table Lens: Merging Graphical and Symbolic Representations in an interactive Focus + Context Visualization for Tabular Information." Proceedings of CHI '94, ACM Conference on Human Factors in Computing Systems, New York, 1994: 318 - 322 and 481 - 482.www.inxight.com Back