Re: [ai-geostats] Using Jenks optimization in ArcMap - further significance tests required?
- Hi Karen,
One question - it sounds like you are plotting a categorical variable? Or do you have counts of that categorical variable?
In any case, when you're dealing with a variable for which you have no particular reason to think it should be randomly distributed (in terms of the function, NOT in space), then I always think some statistic that involves ranking (i.e., low to high) rather than value is a good one - so calculating observed and expected frequencies per area might be a good start, in your case. You possibly also could have spatial autocorrelation, so you might want to look for that (again, not knowing if your variable is continuous or categorical - there are a variety of techniques out there).
But anyway, no, I would not say the Jenk's optimization method, though a useful visual tool, allows you to make any more than hypotheses about the variable's spatial distribution.
Hope this helps,
----- Original Message -----
From: Karen Bellinger Wehner <karen.wehner@...>
Date: Thursday, January 27, 2005 2:13 am
Subject: [ai-geostats] Using Jenks optimization in ArcMap - further significance tests required?
> I am a graduate student doing a GIS archaeology project using
> 8.1, and in need
> of some statistical advice.
> Here's my question, and pardon me as I realize it betrays
> frightening ignorance of statistics!
> Using the quantities-graduated colors option, I have plotted the
> distribution of various artifact categories across 112 excavation
> units within a fifty acre site. I
> have done this in hopes of identifying areas with heightened
> frequencies of specific artifact types, namely those involved in
> production, so as to identify previously unrecognized craft
> areas at the site (which was plowed and as a result, clearly
> features such as buildings have been destroyed - hence looking
> for concentrations of craft-related artifacts in what remains).
> I have selected the Jenks Natural Breaks/Optimization method
> uses automatically in visualizing the plots using the color
> gradations. I understand that Jenks classifies my data into
> categories that maximize the difference between
> categories and minimize the differences within categories.
> So here is the question: Does this process intersect with a chi-
> or goodness of fit test? Does Jenks obviate the need for further
> significance or "goodness of fit" testing? If not, would chi
> be the way to go in assessing the significance of the observed
> artifact distributions?
> ANY advice would be much appreciated.
> Karen Bellinger Wehner
> PhD candidate
> Anthropology Department
> New York University