This is probably a bit of a long shot, but what knowledge lurks in the minds of men? Only the Shadow knows.

An important bit of info, I am using MATLAB with the Mapping Toolbox to do this, I do not have ArcGIS.

Summary question: how can I compute a best fit bounding ellipse for a set of latitude and longitude coordinates? I need the semi-major and semi-major axes measured in degrees or radians, or the semi-major axis and eccentricity.

Long version: I have several sets of geographic data in latitude and longitude coordinates. I want to compare the clustering and distribution of different groupings of these points and display on a map with each grouping displayed inside of a closed polygon containing the points. This isn't hard to do with standard tools if the enclosing polygons are circles. For the most part this works, but one of the data sets has an outlying point that causes the enclosing circle to be much larger than it should be. Of course, a circle drawn with radius computed by the Standard Distance alleviates this issue, but I need my graphs so obvious, even a caveman could understand them. If I have to explain why some points lie outside the circle that is supposed to represent them, I have lost the argument. So, what I want to compute is an ellipse, which will have a long semi-major axis skewed to the outliner, but the semi-minor axis will still be able to show tight clustering of points. That is, a long skinny cigar shape would still give a visual effect of points being clustered near the semi-major axis.

Any ideas? Internet searches have not really turned up anything I recognize as a viable solution. If worse comes to worse, I can setup up my own least-squares solution to this, but that will require some thought into how to setup a Cartesian coordinate system to project the lat/lon coordinate upon.