i.maxlik

GRASS GRASS user
1

A quick one for the statisticians among you: am I right in assuming that
the maximum likelihood classifier can be defined as a form of
discriminant analysis, and if so, does the assumption of equal covariance
matrices of samples apply? A good reference would be appreciated as well!

Cheers,

Philip Verhagen
--

      S t i c h t i n g R A A P

    Regionaal Archeologisch Archiverings Projekt

adress: Plantage Muidergracht 14
          1018 TV Amsterdam
           THE NETHERLANDS
phone: (31) 20 525 5835
fax: (31) 20 525 5834
e-mail: motte@xs4all.nl

2

Take it easy.
1. ""am I right in assuming that
the maximum likelihood classifier can be defined as a form of
discriminant analysis...?""

No way. ML is just a way to measure a distance. Discriminant analysis
implies a (linear) transformaion so that your classes (as defined by your
training fields) produce a partition in which the ratio between the "within
class variances" and the "between classes variances" is maximized.

2. ""...does the assumption of equal covariance
matrices of samples apply? ""
Depends on your data: test it.

Hope it helps!
Agus