Jump to content

 
Photo

mapping clusters of suicides

- - - - -

  • Please log in to reply
15 replies to this topic

#1
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

hi All

I'm working on the project exploring geography of suicide in Wales.. I've been trying to visualize spatio-temporal clusters over three years period..

attached is a first [very] draft version of my map.. I'd like to prepare 9 of them for different time periods where we identified clusters.. and present them using this template as 3x3 matrix..

the audience will primarily have medical background so i'd like to keep it as simple as possible..

and since my cartographic skills are fairly limited I'd be very happy to hear your comments, suggestions and criticism..

thanks a lot

radek

#2
Dennis McClendon

Dennis McClendon

    Hall of Fame

  • Validated Member
  • PipPipPipPipPipPipPip
  • 1,084 posts
  • Gender:Male
  • Location:Chicago
  • Interests:map design, large-scale maps of cities
  • United States

Well, it's a perfectly serviceable map of Wales, with a circle drawn on it.

Did you intend for it to show suicide locations so the clustering would be seen by the audience? If so, at this scale, wouldn't there appear to be a clustering of virtually any human phenomenon where the population density is higher? In other words, whether mapping suicides, musical prodigies, or trainspotters, won't there be a cluster of them in Cardiff?
Dennis McClendon, Chicago CartoGraphics
chicagocarto.com

#3
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

there will be eight more maps..

each with a circle and percentage value..

all referring to different time periods spanning 3years in total..

we have access to pint data.. but needless to say we cannot map them.. so that explains my choice of boundaries of cluster..

another point i wanted to make for the audience is that administrative boundaries should not be taken into account [which is often a case ]..

btw its quite away from Cardiff.. in so called 'valleys' which makes it important.. and some other clusters do not appear in densely populated areas as well..

#4
frax

frax

    Hall of Fame

  • Associate Admin
  • PipPipPipPipPipPipPip
  • 2,320 posts
  • Gender:Male
  • Location:Stockholm, Sweden
  • Interests:music, hiking, friends, nature, photography, traveling. and maps!
  • Sweden

I would be curious in the importance of identifying clusters, and what purpose that has - and also how the temporal resolution works - and if it really makes sense to see these time periods in a discrete fashion. What would it look like if you presented all suicides?
Hugo Ahlenius
Nordpil - custom maps and GIS
http://nordpil.com/
Twitter

#5
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

discrete time boundaries were identified by analysis as well..

tried to present all clusters in the same map.. but not sure how I can add temporal dimension in this case :[

when presenting all suicides - they obviously follow population distribution.. Risk-Adjusted Nearest Neighbour Hierarchical Spatial Clustering done with this package doesn't pick up anything significant..

we think that time is important issue here..

purpose of this exercise is

1. to try answer if we can quickly identify clusters of rare events

2. to serve as first step in identifying areas where further, more qualitative studies may be necessary

#6
Giasen

Giasen

    Contributor

  • Validated Member
  • PipPip
  • 20 posts
  • Gender:Male
  • Location:scottsdale, az
  • United States

discrete time boundaries were identified by analysis as well..

tried to present all clusters in the same map.. but not sure how I can add temporal dimension in this case :[

when presenting all suicides - they obviously follow population distribution.. Risk-Adjusted Nearest Neighbour Hierarchical Spatial Clustering done with this package doesn't pick up anything significant..

we think that time is important issue here..

purpose of this exercise is

1. to try answer if we can quickly identify clusters of rare events

2. to serve as first step in identifying areas where further, more qualitative studies may be necessary


Any chance I could look at the data? I'd like to take a stab at it, if you'll pardon the expression.

Posted Image
GPS

#7
MapMedia

MapMedia

    Hall of Fame

  • Validated Member
  • PipPipPipPipPipPipPip
  • 1,029 posts
  • Gender:Male
  • Location:Davis, California
  • United States

purpose of this exercise is

1. to try answer if we can quickly identify clusters of rare events

2. to serve as first step in identifying areas where further, more qualitative studies may be necessary

Sounds like academic talk - is this for a thesis?
What is your hypothesis, or null hypothesis? I really imagine economics, temporal aspect (season, etc.) playing more of a role than strict geography, which is what the map portrays.
Have you looked at cluster mapping methods used in disease mapping / prediction?

#8
Hans van der Maarel

Hans van der Maarel

    CartoTalk Editor-in-Chief

  • Admin
  • PipPipPipPipPipPipPip
  • 3,898 posts
  • Gender:Male
  • Location:The Netherlands
  • Interests:Cartography, GIS, history, popular science, music.
  • Netherlands

I think it would be useful to map the point locations of *all* suicides, along with population density preferably on a sub-county level.

The thing is, you're *saying* the cluster occurs well outside of the major population centers, in the valley's. But somebody who's not familiar with the layout of Wales will not be able to read that off the map.

I think people from a medical background would be perfectly capable of reading a choropleth map with dots on top of it.
Hans van der Maarel - Cartotalk Editor
Red Geographics
Email: hans@redgeographics.com / Twitter: @redgeographics

#9
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

Any chance I could look at the data? I'd like to take a stab at it, if you'll pardon the expression.


sorry.. i wont be able to reveal any individual data..

i may be able to send you shp files with clusters tho.. but need to double check that..

cheers
radek

#10
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

thx for comments :]

Sounds like academic talk - is this for a thesis?

it's academic work indeed.. it is going to be one of the first outputs of 'An exploratory study of the human geography of suicide and its relationship to place in Wales' project at Swansea University..

I really imagine economics, temporal aspect (season, etc.) playing more of a role than strict geography, which is what the map portrays.

we're trying to find relationships among data we have and see if spatial analysis & GIS may be any help here..

i first tried to identify saptio-temporal clusters in order to look for characteristics of these places.. or specific time periods.. sort of downscaling from the whole area of Wales into regions..

Have you looked at cluster mapping methods used in disease mapping / prediction?

i chose space time permutation.. bcause it doesnt require control population.. and takes into account time dimension [we only have month precsision tho]

#11
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

hi Hans..
thx for feedback..

I think it would be useful to map the point locations of *all* suicides, along with population density preferably on a sub-county level.

i hope to include something like that.. but still waiting for decision of our data provider [Office for National Statistics] which has to appprove all the oputputs..
clusters are tempting and easy to get a clearance from them since they do not include information about individual cases but are still informative..

The thing is, you're *saying* the cluster occurs well outside of the major population centers, in the valley's. But somebody who's not familiar with the layout of Wales will not be able to read that off the map.

I think people from a medical background would be perfectly capable of reading a choropleth map with dots on top of it.

what would be the best option to present it then? i got a raster image with 25m cell presenting population denity.. alternatively i can map figures for small statistical areas..

at first i thought about presenting density of suicides and adding small inset with pop density? not sure if dots on choropleth map map will be readable?

cheers
radek

#12
Hans van der Maarel

Hans van der Maarel

    CartoTalk Editor-in-Chief

  • Admin
  • PipPipPipPipPipPipPip
  • 3,898 posts
  • Gender:Male
  • Location:The Netherlands
  • Interests:Cartography, GIS, history, popular science, music.
  • Netherlands

The size of the symbol makes it represent a fairly large area in the real world. There's no way that somebody can track down a single suicide case just based on that map.

As for the choropleth + dots (or the pop density grid + dots), that would work visually as long as you make sure the symbols and your cluster lines contrast enough. The fact is that if you don't do this, the map really isn't going to convey a lot of information and is verging on being unnecessary.
Hans van der Maarel - Cartotalk Editor
Red Geographics
Email: hans@redgeographics.com / Twitter: @redgeographics

#13
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

The size of the symbol makes it represent a fairly large area in the real world. There's no way that somebody can track down a single suicide case just based on that map.

I fully agree with that.. but our data provider cannot understand it :[

As for the choropleth + dots (or the pop density grid + dots), that would work visually as long as you make sure the symbols and your cluster lines contrast enough. The fact is that if you don't do this, the map really isn't going to convey a lot of information and is verging on being unnecessary.

would that do?
pop density raster
dots [for all suicides in Wales in given period?]
cluster boundary

thx

#14
radek

radek

    Contributor

  • Validated Member
  • PipPip
  • 39 posts
  • Gender:Male
  • Switzerland

update :]
[attachment=1259:final_draft_2.jpg]

#15
dsl

dsl

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 225 posts
  • Gender:Male
  • Location:Denver
  • United States

You may consider using this program http://www.satscan.org/techdoc.html. I believe it is made for this type of analysis, but I have never used it. Scanning through the documentation it seems it can be used without controls and just cases.

Any sort of clustering statistic is subject to scale and boundary. I assume you will address these in your talk.

You may also consider using a Kernel Density Estimation surface, which will help identify hotspots of suicide as opposed to clusters. If you present this with a KDE surface of population, you might be able to demonstrate that suicide hotspots are in different areas from population hotspots. CrimeStat has KDE tools. KDE is dependent on bandwidth selection, but you might consider variable bandwidth for this type of data.

Just my 2 cents. ;)




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users

-->