Mark,
To simply answer your question, no the derived figures do not represent the
number of retailers per square kilometre. The derived figures are just the
area of the Natural Neighbourhood region if they were to be created. These
values are based upon the inverse relationship between the size of natural
neighbourhood region ant the closeness points are to each other. The closer
points are the smaller the regions, the farther apart they are the larger
the region.
For more detail read on...
The traditional way of portraying point density is to divide the study area
up into evenly spaced square areas from which the number of points are
totaled. In this way you will get, for example, 25 customers per square
kilometer. In other words the size of the area is constant and the number
of customers is variable.
x customers per 1 km�
The natural neighbour technique for calculating point density approaches the
problem from a different perspective. This calculation uses the fact that
the closer points are together, the smaller the voronoi regions that will be
created. Conversely, those points that are farther apart produce regions
that are larger. Therefore, it safe to assume that the area of each voronoi
region reflects the degree of density of the points. In this way there is
always one point for every area which becomes the constant and the size of
the area is the variable.
1 customers per x km�
When performing the Point Density operation in Vertical Mapper, a new column
is created in the selected point file. It is updated with the area of each
voronoi region that would be generated around each point. Once this has
been completed, a surface can be generated using the area value attributed
to each point. The difference with the more traditional method is that with
the natural neighbour method, each region has only one point and it is the
size of the region changes, and the more traditional method varies the
number of points while the square area stays constant.
The thing to remember is the updated values are not density values in the
traditional sense however they are representative of density. The dilemma
occurs when you produce a continuous surface from these area values. The
dilemma being the units of the grid , what are they? They are really
representative values that have no units which most people have trouble
accepting.
If you want a more traditional way try using the Square Cell aggregation
technique located in the Point Aggregation With Statistics utility.
Aggregate your points using the area size of your choice (aggregation
distance setting) and make sure you check both the Create Aggregation
Regions setting and the Number of Points Aggregated setting. You can delete
the aggregated point file when you are done because you are not interested
in it. You are however interested in the regions. What you should have is
a bunch of square regions at a specific size with the number of points that
fall inside. Now all you have to do is update a column with the density
(points / area).
I hope this helps.
James Harvie
Vertical Mapper Product Manager
Northwood Geosience Ltd.
43 Auriga Drive, K2E 7Y8
Nepean, Ont, Canada
613-224-2020
-----Original Message-----
From: Rudi van Rensburg <[EMAIL PROTECTED]>
To: MapInfo-L <[EMAIL PROTECTED]>
Date: Friday, June 11, 1999 3:41 AM
Subject: MI Point Densities in Vertical Mapper
>I am in the process of generating trade areas for a selection of
>retailers in various regions of South Africa. I've calculated Point
>Densities using the Natural Neighbour Analysis --> Calculate Point
>Density function of Vertical Mapper. Do the figures thus derived
>represent the number of retailers per square km, which is what I would
>have thought (if my distance units are set to km in my MapInfo). However
>if I Sum all these density values then surely I should get the total
>number of retailers that I used to generate the density values
>initially.
>
>I can find no reference (manual, help etc) to the manner in which these
>densities are derived. Anyone's views?
>
>Mark Lorenz
>
>----------------------------------------------------------------------
>To unsubscribe from this list, send e-mail to [EMAIL PROTECTED] and put
>"unsubscribe MAPINFO-L" in the message body, or contact [EMAIL PROTECTED]
>
>
----------------------------------------------------------------------
To unsubscribe from this list, send e-mail to [EMAIL PROTECTED] and put
"unsubscribe MAPINFO-L" in the message body, or contact [EMAIL PROTECTED]
