Date: Tue, 16 Apr 2002 14:32:48 -0700
Reply-To: "Dennis G. Fisher" <dfisher@CSULB.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: "Dennis G. Fisher" <dfisher@CSULB.EDU>
Subject: Re: maximizing kappa statistics (problems of labelling)
Content-Type: text/plain; charset=us-ascii
You could also try the Adjusted Rand Statistic if you have different number of clusters.
Dennis Fisher
John Uebersax wrote:
> Hi Mike,
>
> You might consider using lambda statistics instead of kappa to compare
> clustering solutions. Advantages include:
>
> 1. Cluster solutions can have different numbers of groups
> 2. No need to match up the clusters in two solutions
>
> You can read up on lambdas in the SAS documentation under SAS PROC
> FREQ Measures of Agreement.
>
> In some cases one might loose a little power by using lambdas instead
> of kappa--because kappa is testing a more specific hypothesis. But on
> the other hand, consider that the matching-up of two solutions is
> usually done by the researcher in a way so as to maximize agreement.
> That would seem to invalidate the p value of kappa--an issue I haven't
> seen addressed in the clustering literature.
>
> All this is avoided by just using lambda statistics.
>
> --------------------------------------------------------------------------------
> John Uebersax, PhD (858) 597-5571
> La Jolla, California (858) 625-0155 (fax)
> email: jsuebersax@yahoo.com
>
> Statistics: http://ourworld.compuserve.com/homepages/jsuebersax/agree.htm
> Psychology: http://members.aol.com/spiritualpsych
> --------------------------------------------------------------------------------
>
> Mike Cheung <mikewlcheung@hongkong.com> wrote in message news:<3CBA9852.9090605@hongkong.com>...
>
> > I am try to calculate the kappa statistics from the output of cluster
> > analysis from two clustering methods...
> > Since the labels of clusters are arbitary, they are not ready for
> > Kappa statistics,
--
Dennis G. Fisher, Ph.D.
Director
Center for Behavioral Research and Services
1090 Atlantic Avenue
Long Beach, CA 90813
562-495-2330
562-983-1421 fax
To him who devotes his life to science, nothing can give more happiness than increasing the number
of discoveries. But his cup of joy is full when the
results of his studies immediately find practical application. There are not two sciences. There
is only one science and the application of science,
and these two activities are linked as the fruit is to the tree.
--Louis Pasteur
| 
