shrug-l: Been a while since I've done this...

Philip Griffith pdgriffith at fsu.edu
Tue Nov 27 16:41:59 EST 2018


Hi Duane,

There have been a lot of good suggestions already. Hopefully here's another!

If by "correlation" you mean something like an observable, mappable pattern that shows up when you put all of the data attributes on a map, you might consider looking at the "Conditional Maps" you can create using GeoDa<https://geodacenter.github.io/>, a free, open source software application by the University of Chicago's Center for Spatial Data Science<https://spatial.uchicago.edu/>, which focuses on exploratory spatial data analysis.

Here's an example of what conditional maps look like using some of the sample data that GeoDa provides:

[cid:image002.jpg at 01D48670.1C908740]

This is from the National Consortium on Violence Research and shows homicide rates per 100,000 in Atlanta, Georgia conditioned on two variables: resource deprivation (or poverty) and police expenditures. Both poverty and the amount spent on police are split into three quantiles (think Low, Medium and High), but they can easily be split up using any other system, from equal intervals to natural breaks to standard deviations.

Here's how you read the output:

The bottom left map only displays counties that have both low poverty rates and low police expenditures. The homicide rates for those counties are then visualized like any other choropleth map. (Here at FREAC we would recommend using the US National Grid to display this data, but I'll leave that for another time. :)) By contrast, the top right map only shows homicide rates for those counties that have both high poverty rates and high police expenditures, the top left map shows low poverty and high police expenditure, etc.

What pops out? At first glance, we can tell from the upper right map that the south side of Atlanta has a mix of high poverty, high police expenditures and higher homicide rates, especially Taylor County (which, incidentally, was still holding segregated proms in the 2000s). Perhaps also interesting is Fulton County in the middle right map: the heart of the City of Atlanta has high poverty, only medium police expenditure, and a very high homicide rate. You'd have to do a lot more investigating to show that the two variables in question were causing the high homicide rate (and not just correlated with it), but at least exploring the data in this way points you in a certain direction.

I hope this makes sense. You could imagine mapping crime rates within census tracts while using the number of street lights and dilapidated housing in those census tracts as the conditional variables. There's a tutorial that walks you through how to create conditional maps (and a whole lot more) in GeoDa here<https://geodacenter.github.io/workbook/3a_mapping/lab3a.html#conditional-map>. I recommend the whole series!

All the best,
Philip Griffith


From: SHRUG-L [mailto:shrug-l-bounces at lists.dep.state.fl.us] On Behalf Of Duane Treadon
Sent: Tuesday, November 27, 2018 9:49 AM
To: 'SHRUG-L at lists.dep.state.fl.us' <SHRUG-L at lists.dep.state.fl.us>
Subject: shrug-l: Been a while since I've done this...

Hi all,

I need a little help shaking the cob webs out of my brain on how to set up an analysis process to look at correlation.  My Chief has asked me to look at crime and see if there is any relationship between it and dilapidated housing and also look at street lighting.  I am lucky in that I have access to all the data needed and it is easy to extract, geocode, and add to our geodatabase for use in ArcMap or ArcPro .  We have crime data for the past several years, our Building Dept. keeps a good record on housing conditions throughout the city, and the Electric Dept. has all street lights mapped out in our GIS already.  I have ArcAdvance and access to all the spatial analysis extensions and tools available as well as ArcPro and its tools/workflows.

What I need help on is deciding what tools to use and how to interpret the results.  My thought is to look at each element as it relates to crime.  Look at crime and dilapidated housing and how it falls on the map. Then crime and street lights in the same way.  Then bring it all together to see if there is a stronger pattern with all three elements.

So what tools should I use?  Any publications or exercises out there that explain the processing of data to determine if there is a correlation.  Any pdfs you can share that explain the tools in ArcMap that I should use.  Are ESRI products not the way to go?  Is there an easier solution for looking at the data?

p.s. as a follow up my Fire Chief would like me to look and see if there is an increase in fire calls that follows the crime.

Thanks for any info you can provide.


Duane 'JT' Treadon, GISP
Director of Geographic Information Systems | City of Thomasville | Information Technology
Office: (229) 227-4023 | Cell: (229) 221-5991
111 Victoria Place | P.O. Box 1540 | Thomasville, GA | 31799
The Power of Community. The Power of Service.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.dep.state.fl.us/pipermail/shrug-l/attachments/20181127/64494956/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.jpg
Type: image/jpeg
Size: 36959 bytes
Desc: image002.jpg
URL: <http://lists.dep.state.fl.us/pipermail/shrug-l/attachments/20181127/64494956/attachment-0001.jpg>


More information about the SHRUG-L mailing list