Re: Under what condition is Mahalanobis distance OPTIMAL?
cramer_at_unc.edu
Date: 08/15/04
- Next message: Alexey: "HELP: White's versus Breusch-Pagan Test for Heteroskedasticity"
- Previous message: Osher Doctorow: "Re: Correlation P<-->(x,y) = P<-->(x,0) + P<-->(0,y) + 2F(x,y) - 1"
- Maybe in reply to: Gus Gassmann: "Re: Under what condition is Mahalanobis distance OPTIMAL?"
- Messages sorted by: [ date ] [ thread ]
Date: 14 Aug 2004 23:25:26 -0400
In sci.stat.math b83503104 <b83503104@yahoo.com> wrote:
: There are so many kinds of distance metrics.
: Is there a way one can show under what conditions the Mahalanobis
: distance is the optimal distance metric?
There is no such thing as optimal
: At least, it is intuitively correct that Mahalanobis distance is
: better than Euclidean distance when the axes have different weighting,
: or when the axes are correlated.
yes if you use appropriate weighting
: Can it be formulated mathematically exactly under which conditions it
: is optimal?
: Thanks
- Next message: Alexey: "HELP: White's versus Breusch-Pagan Test for Heteroskedasticity"
- Previous message: Osher Doctorow: "Re: Correlation P<-->(x,y) = P<-->(x,0) + P<-->(0,y) + 2F(x,y) - 1"
- Maybe in reply to: Gus Gassmann: "Re: Under what condition is Mahalanobis distance OPTIMAL?"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
