CMSC 838b: Information Visualization - Application Presentation

Computer Questions asked on the National Omnibus Survey (Spring 1999)
Data Analysis with Spotfire

Makoto Konishi
February 28, 2001
konishi@cs.umd.edu


Description of the Data

This data is based on the University of Maryland Survey Research Center's Fifth National Omnibus survey. The survey was designed 1000 (1021) interviews at 48 states, using a list-assisted sample, with random selection of one adult respondent within each sample household. The questions are the demographic, computer and internet usage questions asked on the Spring 1999 Omnibus Survey. The items of data are respondent's attribute, where they used the Internet and e-mail, how long they used, and how long they used home computer by the purpose.
University of Maryland's 1999 National Omnibus Survey

The Visualization

 1. Respondent's Profile

I checked a respondent's profile first. The purpose is checking whether neutrality and objectivity without a deviation of data, and observing a respondent's overall state. The color-coding is used for respondent gender: blue for male, red for female.
The data shows there isn't inclined extremely in demographics items, such as age, the household income, state, and a male-and-female ratio, etc. (Figs. 1-5)

1-1 Household Income of 1996 vs. Age. 

X-axis shows the household income of 1996, Y-axis shows age (year of birth). The data of Y-axis is sorted by ascending order as age. The size of a pie chart shows the number of respondents, and shows male-and-female percentage in blue and red. Although female is a little bigger than male, on the whole, it is broadly distributed from 20 years old to 90 years old on each class. (Fig. 1)
Fig.1

1-2  Family Composition

X-axis shows age or the household income, Y-axis shows number of children or marital status respectively.  They are realistic and reasonable distributions altogether. There are many people without a child comparatively with many married people though. (Fig. 2)
Fig.2

1-3 Others 

I observed also others roughly.  These two sets of charts, X-axis shows age or the household income, Y-axis shows respondent's race or school education (the last grade respondents completed)(Fig. 3), employment status or respondent’s state, (Fig. 4) show that the respondents are nationwide and the distribution are realistic and reasonable. The reason for unemployment is the same also. (Fig. 5).)
Fig.3
 
Fig.4

Fig.5

2 . The Internet Use Situation

2-1  Where They Used the Internet? 

Although it was a little unexpected that the Internet use was seldom progressing at school, it should be consider that a 20 or less-year-old respondent is not included. (3D scatter plot; figure 6) The majority or more respondents used the Internet at home or work. It seems that especially used at work is more progress than at home a little. (Fig. 7) Moreover, respondents who used the Internet at work also used at home in many cases. There isn't inclined deviation extremely on the comparison: census region or the last grade respondents completed vs. used the Internet at home. (Dynamically checked in Fig. 8)
Fig.6
 
Fig.7

Fig.8

2-2 Use Time

I explored the Internet use time with several scatter plots based on household income of 1996 vs. age.
I observed comparison of the Internet use time at home and at work (or school). 
Figure 9 shows minutes of used the Internet at work (above) or at home (below) based on income vs. age. Each shows different tendency a little. 
Similarly, observing comparison of e-mail use time, the "used at work" is longer than "used at home" a little. (Fig. 10) It is observable that retired generation also used the Internet and e-mail.

Fig.9
Fig.10

2-3 Use Time by the Purpose at Home

I observed comparison of the home computer use time by the purpose in the same manner. 
The data has eight choices: use time for word process, entertainment, financial plan, educational soft ware, computer organizing (rearrange IT), work related, family acts, free time acts. Figure 11-14 shows minutes of used home computer by the purpose based on income vs. age.

It seems data can be classifiable according to the charts whether it is used for the specific purpose or it is used for a short time widely.
For instance, you can see long time user for computer organizing (rearrange IT) (Fig. 11), work related (Fig. 12), free time acts (Fig. 13), although they didn't use another purpose. It seems that the high-income people often use financial plan software. 
It is observable comparatively that the long time user has one purpose more concretely. 
Moreover, for example, a short time user often has multiple purposes widely. (Fig. 14)  The most of case is less than one hour (Fig. 15). Actually, this respondent used for financial plan, computer organizing, work related. But it seems that there is something to the data of other items in spite of 0.
Totally, Spotfire provides powerful dynamic queries in this case.
 

Fig.11
Fig.12

Fig.13
Fig.14

Fig.15

3.  Conclusion of the Observations

The observed result is regarded overall as near with how people used the Internet and computer whom we are probably thinking in reality every day. The general trend was observable. However, naturally, there is diversity in each use, for example, if it is enabled to divide and analyze who are working about computer specially or not so etc., the analysis will become still more interesting.

Critique for Spotfire

I used Spotfire Pro 4 at this time. I try to mention several comments.

Good Points

  • Dynamic analysis of data that has many parameters or items. 
  • In case data overlapped on a chart or dense, Jittering and Zooming are useful for choice the data very well. (Fig. 16)
  • You can eliminate a potential anomaly or arbitrary points on the chart easily and dynamically. It is strong interaction about more realistic and rational analysis.
  • A data selection can be synchronized in two or more charts at once. It provides relations in the complicated data visually.
Fig.16

Suggestions for an Improvement 

  • Spotfire can't adjust Zoom bar and slider for every step in the chart properties dialog box. It is hard to adjust align axis and the size of marker when I wanted to arrange stacked charts neatly.
  • It seems that a small problem is in the accuracy of the display like figure 15. 
  • I couldn't find out how to set up several charts as same condition on properties dialog box. I think it should have. (I repeated the same operation.)
  • During adjustment dynamic queries, the number of visible data (it is displaying on the bottom end of a window) decreased without clue. I didn't mean to make decrease visible data.
  • In some cases, I wanted to carry out Undo operation. It happened several times. I think every soft ware should have Undo.
  • Export charts were different aspect ratio from what I created and seen in the screen. I think the exported charts should be as same appearance as simple screen shot. Because users might get confuse.
  • I couldn't find out how to display average value quickly. I guess Spotfire probably has it, but it was unclear for me.)

Makoto Konishi
Last modified: Wed Feb 28 2001