[Resonding to a question on grass-web]
---------- Forwarded message ----------
From: rosa.coluzzi <rosa.coluzzi@imaa.cnr.it>
Date: Wed, Feb 22, 2017 at 6:11 PM
Subject: [GRASS-web] r.kappa help
To: grass-web@lists.osgeo.orgGreetings,
please, can you help me to understand what's exactly the "Estimated
Kappa value" in output of r.kappa routine?
The kappa coefficient measures the agreement between two classifications by taking the observed agreement between the two classification and subtracting from that the probability that the agreement is only by chance.
Normally, this is calculated across all classes and thus is not conditional to which classification you chose as reference.
The 'Estimated kappa value' in r.kappa is the value only for one class, i.e. the observed agreement between the classifications for those observations that have been classified by classifier 1 into the class i. In other words, here the choice of reference is important.
It is calculated as (taken from the code [1]):
kpp[i] = (pii[i] - pi[i] * pj[i]) / (pi[i] - pi[i] * pj[i]);
where=
pii[i] is the probability of agreeement (i.e. number of pixels for which there is agreement divided by total number of assessed pixels)
Pi[i] is the probability of classification i having classified the point as i
Pj[i] is the probability of classification j having classified the point as i
Hope this helps.
Moritz
[1] https://trac.osgeo.org/grass/browser/grass/branches/releasebranch_7_2/raster/r.kappa/calc_kappa.c