I have a population of objects (each of which is described by a list of about a half-dozen real numbers), and I'd like to find some reasonable kind of distance between each of several different test objects and the median of the population, with the goal of being able to say, for example, that because test object A is closer to the median than test object B, therefore A is "more like" the population than B.
I believe that I can use the
MultivariateStatistics package to characterize my population, and then either
SimplexMedian to find a center. So then I need to find a measure of distance from that center to each test object. I gather that a simple
EuclideanDistance would be incorrect, because I need to account for the ellipsoid's size and orientation (I believe, but I'm not certain, that this is the Mahalanobis distance). I don't see a function in the
MultivariateStatistics package that lets me find such distances.
Is it possible to compute a meaningful distance from a test object to the median of the population that will let me assert whether one test object is "more like" the population than another, and if so, how do I compute that distance? Practical advice would be much better than theoretical. Thanks!