[Openspace] quantile plot

Dan Brown danbrown at umich.edu
Mon Sep 20 19:47:39 CDT 2004

I don't think this addresses the problem, or maybe only partially.  
Under some circumstances, and I haven't investigated to know exactly 
which, GeoDa puts two times the appropriate number of spatial units in 
the second quantile and none in the first quantile (look at the numbers 
in the categories in the example Michael provided).  We had the same 
problem as I was using GeoDa in a class.  It seems this might merit 
investigation as a possible bug.  In the case we were looking at, I'm 
sure there wasn't a tie at the low end that had two-times the target 
number of spatial units in the quantiles.

Julia Koschinsky wrote:

>One of the reasons why you can end up with different numbers of observations
>in your quintile categories is if you are using count data that is skewed,
>for instance has a lot of zeros or small numbers (as is the case with
>certain mortality or risk data). In those cases, the computation of break
>points is unreliable because there are ties between the frequently occurring
>values. The use of rates instead of counts is more appropriate here.
>Please write back if this does not address your problem.
>Julia Koschinsky
>Spatial Analysis Laboratory
>University of Illinois at Urbana-Champaign
>-----Original Message-----
>From: openspace-bounces at agec221.agecon.uiuc.edu
>Behalf Of Michael Anyadike-Danes
>Sent: Monday, September 20, 2004 10:32 AM
>To: openspace at agec221.agecon.uiuc.edu
>Subject: [Openspace] quantile plot
>I have a dataset with 434 observations.
>I requested a quantile map with 5 divisions (ie a quintile map).
>The map has five divisions: "1st range (0); 2nd range (165); 3rd range (87);
>4th range (95); 5th range (87)".
>I was expecting 5 more-or-less equal ranges (86 and/or 87).
>Have I missed something?
>Michael Anyadike-Danes
>Economic Research Institute of Northern Ireland
>Openspace mailing list
>Openspace at sal.agecon.uiuc.edu

More information about the Openspace mailing list