An improved Fuzzy Kappa statistic that accounts for spatial autocorrelation

Hagen-Zanker, A (2009) An improved Fuzzy Kappa statistic that accounts for spatial autocorrelation International Journal of Geographical Information Science, 23. pp. 61-73.

	Text (deleted) FuzzyKappa_2009.pdf Restricted to Repository staff only Available under License : See the attached licence file. Download (545kB)
Preview	Text final_submitted.pdf Available under License : See the attached licence file. Download (640kB)
Preview	Text (licence) SRI_deposit_agreement.pdf Download (33kB)

Abstract

The Fuzzy Kappa statistic expresses the agreement between two categorical raster maps. The statistic goes beyond cell-by-cell comparison and gives partial credit to cells based on the categories found in the neighborhood. When matching categories are found at shorter distances the agreement is higher. Like the well-established Kappa statistic the Fuzzy Kappa statistic expresses the mean agreement relative to the expected agreement. The model underlying the expected agreement assumes absence of spatial autocorrelation in both compared maps. In reality however, spatial autocorrelation does lower the expected agreement as matching categories become less likely to be found close-by. Since most maps have some degree of spatial autocorrelation, the calculated expected agreement is generally higher than the true expected agreement. This leads to counterintuitive results when maps that appear to have considerable agreement obtain negative Fuzzy Kappa values. Furthermore, the Fuzzy Kappa may be biased, as it systematically attributes lower agreement to maps with stronger spatial autocorrelation. This paper proposes an improved Fuzzy Kappa statistic that is based on the same local agreement and has the same attractive properties as the original Fuzzy Kappa. The novelty is that the new statistic accounts for spatial autocorrelation, such that the expected Fuzzy Kappa for maps that are not cross-correlated is equal to zero. The improved statistic is applied on two cases to demonstrate its properties.

Item Type:	Article
Divisions :	Faculty of Engineering and Physical Sciences > Civil and Environmental Engineering
Authors :	Name Email ORCID Hagen-Zanker, A
Date :	2009
Identification Number :	10.1080/13658810802570317
Uncontrolled Keywords :	fuzzy kappa map comparison accuracy validation classification accuracy landscape analysis categorical maps algorithms agreement patterns error
Additional Information :	This is an Author's Accepted Manuscript of an article published in International Journal of Geographical Information Science, 2008, copyright Taylor & Francis, available online at: http://www.tandfonline.com/10.1080/13658810802570317
Depositing User :	Symplectic Elements
Date Deposited :	15 Aug 2013 09:17
Last Modified :	31 Oct 2017 15:07
URI:	http://epubs.surrey.ac.uk/id/eprint/790291

Actions (login required)

View Item

Downloads