[GRASSLIST:5246] Re: Local density analysis script for GRASS 5.7

[Michael Barton]

Benjamin,

Here are a some references.

Johnson, I. (1984). Cell frequency recording and analysis of artifact
distributions. Intrasite Spatial Analysis in Archaeology. H. J. Hietala.
Cambridge, Cambridge University Press: 75-96. [I referenced this one in the
script header]

Johnson, I. (1995). Mapping archaeological data: a structured introduction
to MapInfo. Sydney, Archaeological Computing Laboratory Sydney University.
[The algorithm I used is based on one that Johnson describes here]

Kintigh, K. W. (1990). Intrasite spatial analysis: a commentary on major
methods. Mathematics and Information Science in Archaeology: a Flexible
Framework. A. Voorrips. Bonn, Holos: 165-200.

Michael


On 1/3/05 2:57 AM, "Benjamin Ducke" <benducke at compuserve.de> wrote:

> Sounds interesting. I have some site maps here that I would like
> to test it on.
> Could you give me some lit reference for Ian Johnson's LDA algorithm?
> 
> Cheers,
> 
> Benjamin
> 
> On Sun, 02 Jan 2005 18:31:28 -0700
> Michael Barton <michael.barton at asu.edu> wrote:
> 
>> Here is something I've wanted to do for awhile.
>> 
>> This script calculates Ian Johnson's Local Density Analysis for vector
>> points. It is used in archaeology to examine the degree of clustering at
>> varying neighborhood radii of sites or artifacts or the degree of clustering
>> of one artifact class in the vicinity of another artifact class. It is a
>> nice routine, but rarely available.
>> 
>> I've been futzing with this for awhile in GRASS and finally worked out how
>> to do it. The problem was determining the number of points within a given
>> radius of a point. This should not be difficult in a GIS like GRASS, but it
>> is surprisingly tricky. R.neighbors doesn't have a count option and hacking
>> it by giving all relevant cells a cat of one and doing a sum doesn't give a
>> correct answer. Buffering also doesn't give the same answer as summing the
>> count of point around each point as the latter can count some points twice
>> (correctly), whereas buffering only will count them once. V.neighbors
>> produces an output that is NOT a count of points around each point but
>> something quite different. Anyway, I was able to use v.distance with the -a
>> and -p flags to get this. This method may be useful for other kinds of
>> similar clustering or neighborhood analyses.
>> 
>> For simply looking at the degree of clustering in a points file, use the
>> same file for the base and analysis files. For looking at the degree of
>> clustering for points of type 2 in the neighborhood of points of type 1, use
>> type 1 for the base file and type 2 for the analysis file.
>> 
>> You can select minimum and maximum neighborhood radii, along with a value to
>> increment the neighborhood radius from minimum to maximum.
>> 
>> The main output is in a text file, to which you can append a series of
>> analyses. For a single analysis, you can also have a line graph of the
>> results displayed.
>> 
>> I've used this kind LDA with an algorithm described by Ian Johnson for
>> MapInfo and in Keith Kintigh's archaeological analysis program. The results
>> look correct here. But I'd appreciate it if anyone cares to test it and let
>> me know what they think.
>> 
>> I'll post this on my website in the script exchange area.
>> 
>> Cheers and Happy New Year
>> Michael Barton
>> ____________________
>> C. Michael Barton, Professor of Anthropology
>> School of Human Evolution and Social Change
>> PO Box 872402
>> Arizona State University
>> Tempe, AZ  85287-2402
>> USA
>> 
>> Phone: 480-965-6262
>> Fax: 480-965-7671
>> www: <www.public.asu.edu/~cmbarton>
>> 
>> 

____________________
C. Michael Barton, Professor of Anthropology
School of Human Evolution and Social Change
PO Box 872402
Arizona State University
Tempe, AZ  85287-2402
USA

Phone: 480-965-6262
Fax: 480-965-7671
www: <www.public.asu.edu/~cmbarton>