I
have three bits of data and I want to see if there's a correlation
between the three. The three data topics are all for parcels in the
township I work for.
The first is a house and site condition
survey that was conducted. Each residential property was rated on its
condition for various features and an average condition was given. This
average is between 1 and 4, with 4 being an excellent condition.
The
second bit of data is the number of service calls (police, fire, and
EMS) to the property in the last 6 years. Each parcel simply has the
number of calls to that location.
The third bit of data is if
the residential property is renter or owner occupied. Each residential
property has a field that simply says Renter or Owner.
Now,
what I would like to do is try and find a correlation between these
three pieces of information and be able to answer questions like, does
renter occupied housing have a higher crime rate, or does housing
condition have anything to do with crime rate?
Does anyone have
an idea of how I could get started doing this? I haven't had a
statistics class in quite some time and have never undertaken such a
task. Any thoughts or ideas are appreciated. Thank you!
Eric
Spatial Correlation Mapping
Started by
cmdrico7812
, Apr 06 2006 01:24 PM
2 replies to this topic
#1
Posted 06 April 2006 - 01:24 PM
#2
Posted 06 April 2006 - 02:37 PM
Multiple
regression will get you your relationships not accounting for space,
you need to identify your independent & dependent variables. If you
want space as a component then you will need to look at spatial
statistics and look at spatial patterns or hotspots in one variable in
space and see if there are any relationships in those pockets with the
other variables... I'm no expert either so these are just ideas of the
top of my head as to how I would start thinking about this. Some of the
GIS list servers and a good book on crime mapping and spatial
statistics might be a good start.
In terms of displaying the
results as a map it's hard to say without looking at the results of the
analysis but its essentially a multivariate problem and you can
approach it from a small multiples perspective (preferable) in which
you display relationships betwen two variables at once or a a three
variable multivariate map (more complicated and difficult to
communicate effectively)...
regression will get you your relationships not accounting for space,
you need to identify your independent & dependent variables. If you
want space as a component then you will need to look at spatial
statistics and look at spatial patterns or hotspots in one variable in
space and see if there are any relationships in those pockets with the
other variables... I'm no expert either so these are just ideas of the
top of my head as to how I would start thinking about this. Some of the
GIS list servers and a good book on crime mapping and spatial
statistics might be a good start.
In terms of displaying the
results as a map it's hard to say without looking at the results of the
analysis but its essentially a multivariate problem and you can
approach it from a small multiples perspective (preferable) in which
you display relationships betwen two variables at once or a a three
variable multivariate map (more complicated and difficult to
communicate effectively)...
#3
Posted 13 April 2006 - 12:50 PM
"I
assume you're on ARCGis Platform? If you're on 9.1, there is a toolbox
called ""Spatial Statistics"". In it are hot spot identifiers, clusters,
and spatial correlation and more.
I would try and look in the toolbox and start to see if there are scripts already made up for what you're trying to do.
HTH "
assume you're on ARCGis Platform? If you're on 9.1, there is a toolbox
called ""Spatial Statistics"". In it are hot spot identifiers, clusters,
and spatial correlation and more.
I would try and look in the toolbox and start to see if there are scripts already made up for what you're trying to do.
HTH "
Randy Long
GIS/CAD Tech
Mackay and Somps
Raster is Faster, but Vector is Corrector.
GIS/CAD Tech
Mackay and Somps
Raster is Faster, but Vector is Corrector.
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users


Sign In
Create Account

United States
Back to top










