Hi! Working on a litle datebase (903 lines) I notice a difference between the result of classInterval(style="fisher") and the natural breaks proposed by ArcView GIS (3.2). I read that ArcView works on a sample of data. Is it right ? May it cause a significative difference ? Greetings from Strasbourg, France. Herv? Polesi
classInterval(style="fisher") vs. ArcView GIS natural breaks
3 messages · Hervé Polesi, Roger Bivand, White.Denis@epamail.epa.gov
On Fri, 16 Mar 2007, Herv? Polesi wrote:
Hi! Working on a litle datebase (903 lines) I notice a difference between the result of classInterval(style="fisher") and the natural breaks proposed by ArcView GIS (3.2). I read that ArcView works on a sample of data. Is it right ? May it cause a significative difference ?
The classInt package is open source, but ArcView/ArcGIS isn't (unless you consider the new python scripts, but I don't think these include class intervals). Is it possible to check in the ESRI documentation? I recall being told that the symbology is built on the fly and probably stored in the project, and that a sample is used instead of the total. If anyone has some students who'd like to deconstruct the ESRI approach by testing on known distributions, that might be fun. We have the Fisher code (as implemented), so seeing why things might differ ought to be possible. Of course, there are very many different classifications anyway, even for moderate n and k. Best wishes, Roger
Greetings from Strasbourg, France. Herv? Polesi
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Roger Bivand Economic Geography Section, Department of Economics, Norwegian School of Economics and Business Administration, Helleveien 30, N-5045 Bergen, Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43 e-mail: Roger.Bivand at nhh.no
Just found out a little about this in ArcGIS (ArcMap), version 9.1. Yes, apparently with large datasets, the "natural breaks" classification (and maybe others like quantile) does sample. However, it is possible to change the sample size by clicking the "Sampling" button in the Classification window. With 40,000 cases it was really slow, however. For R users, Roger Bivand's response is more relevant. r-sig-geo-bounces at stat.math.ethz.ch wrote on 2007-03-16 04:08:42:
Hi! Working on a litle datebase (903 lines) I notice a difference between the result of classInterval(style="fisher") and the natural breaks proposed by ArcView GIS (3.2). I read that ArcView works on a sample of data. Is it right ? May it cause a significative difference ? Greetings from Strasbourg, France. Herv? Polesi
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-geo