Koen, Congalton and Green 1999 ("Assessing the Accuracy of Remotely Sensed Data: Principles and Practices") describes such a method which they term Margfit. This method "can be applied to normalize or standardize the error matrices for comparison purposes. Margfit uses an iterative proportional fitting procedure which forces each row and column (i.e., marginal) in the matrix to sum to a predetermined value; hence the name Margfit. If the predetermined value is one, then each cell value is a proportion of one and can easily be multiplied by 100 to represent percentages. The predetermined value could also be set to 100 to obtain percentages directly or to any other value the analyst chooses. In this normalization process, differences in sample sizes used to generate the matrices are eliminated and therefore, individual cell values within the matrix are directly comparable." (Chapter 5).
Dr. Congalton has sent me his Margfit application (I think it is written in C++ but can't be sure) and I'm sure would do the same for anyone else interested. He can be reached at russ.congal...@unh.edu AS for R, I bet it would be relatively straightforward to code a version of Margefit into R. Hope this helps, Andrew Andrew Niccolai Doctoral Candidate Yale School of Forestry 860-402-6079 andrew.nicco...@yale.edu -----Original Message----- From: r-sig-geo-boun...@stat.math.ethz.ch [mailto:r-sig-geo-boun...@stat.math.ethz.ch] On Behalf Of Koen Hufkens Sent: Monday, March 09, 2009 6:19 AM To: R-sig-GEO Subject: [R-sig-Geo] comparing confusion matrices Dear list, I need some expert assistance on a geo-statistical question. I have two maps, first and second choice classification results. And I have two areas selected within this map. For these areas I calculated confusion matrices. However, I want to compare these matrices. I was considering a Mantel test. However, I do not know if this is correct (sample sizes are not the same so I normalized both matrices) and ideally I would love to extract inter row differences rather than the evaluation of the total matrix (Z-score p-value etc). Has anyone experience with these kind of statistics (row based comparison of matrices). Kind regards, Koen _______________________________________________ R-sig-Geo mailing list R-sig-Geo@stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-geo _______________________________________________ R-sig-Geo mailing list R-sig-Geo@stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-geo