Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



Help, Mahalanbois distance
Posted:
Dec 9, 1996 1:06 PM


I read a few multivariate statistics books for Mahalanbois distance, the formula for two distributions is
(v  u) (covariance matrix)^1 (v  u)
where v and u are the mean of two distributions respectively
Someone told me that the Mahalanbois distance is the weighted Euclidean distance with variance. Is this the closest explaination and what is the inverse covariance matrix for?
If I want to calculate the Mahalanbois distance between a ndimensional pt, x', and a distribution, is the following calculation correct?
(v  x') (covariance matrix)^1 (v  x') where v is the mean of the distribution
Thanks in advance Joe



