Title:  

	Remarks on Outliers and Smoothing Conditional Maps


Authors:

	David W. Scott and Blair Christian



Abstract:

	The integration of a full-featured statistical
	system and exploratory visualization tool with
	a geographical information system has demonstrated
	and proven effective for understanding the patterns
	in geographical data.  In this talk, we examine
	the use of robust regression via minimum distance
	estimation in this context.  Analyses of spatial
	data are highly susceptible to anomalous data,
	outliers, and incomplete model specification.
	Diagnostic procedures after-the-fit are often
	difficult to decode.  In contrast, the minimum
	distance fits provide diagnostic information
	directly, which can be explored in a GIS context.


Keywords:  Robust estimation, geographical information systems,
	   regression diagnostics, data visualization.