Dynamic Linking of Arc View, XGobi and XploRe for Multivariate Spatial Data: Linked Brushing for Points, Polygons, and Lines
Nicholas J. Lewin-koh*, Jiigen Symanzik and Dianne Cook
Graphics Lab
School of Computing
National University of Singapore
Lower Kent Ridge Road
Singapore 119260
Tel: (65) 874-6253 Fax: (65) 872-3919
E-mail: kohnicho@comp.nus.edu.sg
Abstract
In this paper we present a dynamic link between Arc View, a desktop GIS, XGopi, a visualization program for multivariate data with well developed graphics tools, and XploRe, a statistical computing environment. Through a Remote Procdure Call mechanism we pass data dynamically among the three packages, thus enabling dynamic brushing to identify spatial outliers, to assist in clustering, and to apply various smoothers to the data for viewing it in relation to the map surface. The different views and tools supported by these packages provide unique insights into data when combined with the added layer of dynamic brushing. We show how we apply brushing in our implementation to point and polygon data, and briefly discuss linear data. We demonstrate how these tools give added insight into epidemiological data from health administration units.
1 Introduction
Spatial data set require additional tools for exploratory data analysis that are not always present in current Geographic Information Systems (GIS). Deficits in GIS become obvious especially when the data are multivariate and the dependencies within and between variables are of interest. There are, however, well developed techniques for the exploration of multivariate data such as the grand tour (Asimov 1985, Buja and Asimov 1986), scatter-plot matrices (Chambers et al. 1983) , and parallel coordinate plots (Inselberg 1985, Wegman 1990). However, spatial dependency cannot always be adequately visualized through these methods alone. In combination with these multivariate views of the data, dynamic maps play an important role in helping to visualize the nature of spatial dependence. With this in mind, we have developed the ArcView / Xgobi/XploRe environment. This is an environment that links three different kinds of software: a GIS, ArcView, a dynamic statistical graphics program. Xgobi (Swayne et al. 1998),
and a statistical computing environment, XploRe (Hardle et al. 1995). The ArcView/Xgobi/XploRe environment allows the exchange of data and commands and provides analytic and graphical methods for the analysis of spatially referenced data in its geographic context. This environment has been successfully used for the analysis of precision agricultural data (Symanzik et al. 1998b), real estate data (Symanzik et al. 1997), and satellite images (Symanzik et al. 1998a). The linked software environment is available for major UNIX workstations such as Sun/Sparc, DEC alpha, and SGI. Additional examples that highlights the ArcView/XGobi component of this environment can be found in Cook et al. (1996) and Cook et al. (1997). A description of the technical components that form the basis of this environment can be found in Symanzik et al. (1999).
This paper focuses on modifications made to the ArcView side of the ArcView/Xgobi/XploRe environment to allow handling of data other than points and to elaborate on how the linked environment exploits the strengths of each package. We extend the functionality of the link to handle polygon and line data beyond the original capabilities to analyze only data at point locations. We use data on cancer mortality and economic variables from health services areas to illustrate our points. Due to space limitations and the importance of color graphics we are posting an extended set of figures at http:/www.public.iastate.edu/arcviewxgobi/ACRS/extended.html.
2 Point, Line and Polygon Data
In the earlier versions of the ArcView/Xgobi/XploRe environment only point data could be handled. Brushing in the ArcView map view occurred by changing the color, size, and type of the glyphs to match those used in XGobi and XploRe. Different groupings in the data are identified by different glyphs. In ArcView, this is done redrawing the points as graphics and manipulating graphics in the internal graphics list. Point data has fairly well defined metrics. Usually, distance is the Euclidean distance between two points, though other metrics also exist. The variogram cloud, multivariate variogram cloud, and the lagged scatter plot all use distance between objects as the basis of further calculations. When initiating the link, the ArcView script that calls XGobi first extracts the x and y coordinates of each data point and passes these coordinates as the first two columns of the data into XGobi. Then XGobi calculates the necessary values for the distance-based statistics from the data that has been passed from
ArcView. The coordinates remain invisible to the user.
With only the capability of using points, it is still possible to examine data based on linear or a real tessellations. Polygons or lines can be converted to point layers based on the center points of the objects. Unfortunately, this approach does not provide a satisfactory view on the map view of the spatial objects. Information is lost regarding the topology of the map that one can see if the objects retain their original shapes. We follow with our approaches how to better handle polygons and lines.
In the most recent version of the ArcView/Xgobi/XploRe environment we added the additional capability to work with polygon and linear features. With point data, all features on the layer of interests are redrawn as graphics in the map view. Brushing takes place by manipulating objects in the view's graphic list. We use the same approach for polygons and lines. We redrawn polygons and lines as graphic objects on the screen. Brushing is accomplished by changing the fill color when objects are brushed. The data sent to XGobi, however, are still the coordinates of the polygon centroid or the center point of the line segment.
With polygons and lines, we run into some conceptual issues due to our representation in the data space of our a real units as points. In the case of point data , there is a one-to-one correspondence between the graphic representation in the data space and map space which allows us to manipulate glyphs and colors simultaneously. With polygons and lines, it becomes harder to match a glyph in the data view with a glyph in the map view if the polygon or line changes to a matching color. Currently, glyph brushing is disabled for polygon and line data types.
