KEYWORDS:
Animation, map, statistics, graphic, query, correlations, atlas,
time trend, health.
ABSTRACT
Dynamic queries are user-controlled animated displays of visual
or textual data. On an application developed for the National
Center for Health Statistics and running on their ordinary staff
equipment, a thematic map of the United States is animated by
moving a time slider to illustrate trends of mortality rates.
Other sliders filter parts of the maps interactively according
to parameters such as demographics. During the filtering possible
correlations can be observed as they create color patterns in
the "dissapearance" of the states. This application
has been received with enthusiasm by the user community and will
be distributed with selected new datasets.
BACKGROUND
Dynamic queries
Dynamic queries (DQs) are user-controlled animated displays of
visual or textual data. Several experiments have demonstrated
that dynamic queries help users find trends and exception conditions
[1, 2]. The National Center for Health Statistics (NCHS) proposed
to investigate the use of dynamic queries as a health statistics
exploration tool. The resulting application will be made available
to researchers and epidemiologists with the release of selected
new datasets.
This tool can be used to browse unexplored data and also to illustrate
findings. For this application [4] the benefits of dynamic queries
for exploring time trends and providing dynamic filtering in the
search for correlations are demonstrated. The feasability of dynamic
queries on thematic maps, even on a 386 PC (the only equipment
widely available to the potential user community) is also seen.
Displaying spatio-temporal statistics
Traditionally, health statisticsólike many other statisticsóare
made available through tables of averages, graphs and pages of
detail data. Printed maps and atlases are used but access to detail
is lost since all values are split by only a few categories used
for the color coding; users still must refer to tables. Geographical
Information Systems are now reaching the desktop world but they
remain aimed at the production of maps rather than the
exploration and rapid browsing of maps.
Showing time trends
For maps, time trends are traditionally shown as a series of small
maps. Some software packages allow canned animations to be prepared
and shown to illustrate findings. Real-time animation is repeatedly
cited in the literature (e.g. [3]) but only as a desirable future
development or reserved for high-end equipment. We believe
that giving users control of the animation is a determining factor
of the application's success. (Actually the video itself cannot
quite convey the sense of control since it is the equivalent of
a canned animation program.)
GENERAL DESCRIPTION OF THE INTERFACES
After the selection of a cause of death and of three parameters
representing possible demographic confounders, a map of the US
is displayed with the chosen mortality rate color coded for each
state (e.g. the Cervix Uteri cancer rate [figure 1]). The states
where the cancer rate is high are red (dark on the figure), the
states where the rates are low appear blue and the other states
are gray. On the right of the screen four sliders provide control
over the map animation.
Showing time trends
When the year slider is moved from 1950 to 1970 the color of each
state changes in real time, reflecting each state's rate change.
Our example shows the overall decline in cervix uteri cancer rates
over the years as well as geographical clusters.
Querying
In our example the three chosen query parameters for each state
are the percentage of the population who went to college, the
per-capita income and the percentage of smokers. The user may
set upper and lower bounds on the range sliders. States with values
that fall outside the boundaries are set to the background color
(black) while states with values within all the boundaries are
set to a color determined by their corresponding mortality rate.
For example, users could look at states having a college education
percentage between 15 and 20%, an average per-capita income greater
than $12,000 and a high percentage of smokers. This technique
allows even first-time users to specify queries easily and with
no errors.
Interactive filtering
When the college slider is slid from 28% (the maximum) to 10%
(the minimum), the states will progressively disappear (turn black)
from the map under user control. States with a high percentage
of college educated residents will be turned off first while the
states with a lower percentage will remain on the map longer.
In the case of cervix cancer it can be seen that the low cancer
rate states tend to disappear first and mainly "reddish"
high rate states persist, suggesting a possible correlation. Because
of the direct control, the animation can be easily replayed at
will and at any speed until a suspected pattern can be isolated.
Access to details
Finally, an obvious but greatly appreciated benefit (compared
to the traditional atlases) is the immediate access to all the
detail data which appears on a form when an area is selected.
FAST ANIMATION
For dynamic queries real-time is essential, which means that changing
the colors on the map must be instantaneous. The traditional technique
of flood-filling regions is known to be too slow; the "states"
version of our system illustrates that the simple but powerful
palette manipulation technique allows dynamic queries to be used
on common machines (e.g., on a 386 PC).
With a greater number of irregularly shaped areas, it is the display
update speed which becomes the challenge for low-power PCs and
not the data access. For example our 800 "county-like"
areas version will not refresh fast enough without efficient animation
algorithms or the replacement of the irregular shapes by small
regular shapes which can be processed faster. Of course zooming
on a section of the map is always possible, making the real time
animation feasible again.
CONCLUSIONS
This application has been very well-received by the user community.
It is a compelling tool for both novices and experts. These methods
hold much promise for improving access to large data sets because
they utilize the remarkable human capabilities for visual pattern
recognition. We want to encourage designers and developers to
consider the use of dynamic queries as there are many practical
applications today which can be implemented on ordinary equipment.
ACKNOWLEDGEMENTS
Partial support was provided by the National Center for Health
Statistics and Johnson Controls. We want to thank Linda Pickle
from NCHS, Boonteck Kuah who worked on the first prototype, and
the team at HCIL.
REFERENCES
1. Ahlberg, C., Williamson, C., Shneiderman, B. Dynamic queries
for information exploration: an implementation and evaluation
ACM CHI ë92 Conference Proc. (Monterey, CA, May 3-7,
1992) 619-626.
2. Williamson, C., Shneiderman, B. The dynamic HomeFinder:
evaluating dynamic queries in a real-estate information exploration
system Proc ACM SIGIR ë92 (Copenhagen June 21-24,
1992) 338-346.
3. Cleveland, W., Ed. Research in Statistical Graphics,
Journal of the American Statistical Association. June 1987,
Vol. 82, No. 398, Statistical Graphics.
4. Plaisant, C. Facilitating data exploration: dynamic
queries on a health statistics atlas. In Proceedings of the
American Statistical Association 153rd Annual Meeting (San
Francisco 1993), Section on Government Statistics. A.S.A. Alexandria,
VA.
Cervix cancer rates are color coded on the map. The year slider
shows time trends. The other sliders allow interactive filtering
of the map according to the three chosen demographic parameters.
Permission to copy without fee all or part of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the ACM copyright notice and the title of the publication anits date appear, and notice is given that copying is by permissionof the Association for Computing Machinery. To copy otherwise, or to republish, requires a fee and/ or specific permaission. CHI94 Companion-4/94 Boston, MA USA