[18950] in s-news-athena
[S] Error matrix and analysis
daemon@ATHENA.MIT.EDU (Steve Bousquin)
Mon Aug 16 10:26:52 1999
Message-Id: <37B81DF1.5642A3B@sfwmd.gov>
Date: Mon, 16 Aug 1999 10:19:29 -0400
From: "Steve Bousquin" <sbousqu@sfwmd.gov>
Reply-To: sbousqu@sfwmd.gov
Mime-Version: 1.0
To: "s-news@wubios.wustl.edu" <s-news@wubios.wustl.edu>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
There is a method of analysis used in remote sensing in
which a reference set of classification data are compared to
another data set (Congalton. 1999. Assessing the accuracy of
remotely sensed data.). In both data sets, items have been
placed into discrete classification categories. The
product is a square "error matrix" that gives the reference
classification in the columns and the other classfication in
the rows, where the columns and rows are the categories in
the classification. The diagonal elements of the matrix are
the numbers of times the 2 data sets agree. Nondiagonal
elements give the numbers of misclassified items by
category.
My question is how I can produce such a table in S-Plus
(v.4.5) using predefined functions, or whether someone has
written functions for this kind of analysis? The input data
table as I see it would consist of columns for the 2 data
sets for a large number of classified items in the rows of
the table. The data would of course be categorical (letter
codes for about 40 categories).
I am also looking for functions to analyze tables of this
kind statistically using the multinomial distribution.
=======================================
Steve Bousquin
Kissimmee River Restoration Division
South Florida Water Management District
West Palm Beach, FL
sbousqu@sfwmd.gov
=======================================
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
