11 replies to this topic

[Pyzdra]

Dear List Members,
We are preparing a population health atlas (different from the one I wrote some time ago on this list) and an issue aroused recently regarding class design in choropleth density maps for various types of urban services. On one hand it would be nice to compare densities of such different services, e.g. No of doctors per 10000, diabetes per 10000, recreation centres per 10000 etc., on the other hand the ranges of various variables are expectedly quite different, so the dilemma of standardized maps vs. informative maps with unique classes is at hand. The cross-map/chapter comparability is somewhat desired, but not crucial. Any recommendations?

[frax]

are you asking if you should use some fixed class breaks, or adapt them to the distribution on a case-by-case basis? If so I would strongly vote the latter.

[Rob]

i'm too not clear if your question is about data normalization/standardization or classification. could you clarify?

thanks,

rob

[Martin Gamache]

Set your class breaks based on the data range/distribution for each map. Although there may be relationships between factors shown on the maps, using the same class breaks for them is not the way to explore them.

[Pyzdra]

Hi Rob,
Sorry for not being too clear. The question is about creating fixed classes for a number of different density maps showing different variables, so they can be easily comparred vs. creating unique classes for each variable based on their distribution. I am seeing from other replies that the latter option is the preferred one.
Thanks Rob.
Peter

i'm too not clear if your question is about data normalization/standardization or classification.  could you clarify?

thanks,

rob

<{POST_SNAPBACK}>

[Pyzdra]

Thanks Martin!
Peter

Set your class breaks based on the data range/distribution for each map. Although there may be relationships between factors shown on the maps, using the same class breaks for them is not the way to explore them.

<{POST_SNAPBACK}>

[Pyzdra]

Thank you Frax - your answer was bang on despite my convoluted question.

are you asking if you should use some fixed class breaks, or adapt them to the distribution on a case-by-case basis? If so I would strongly vote the latter.

<{POST_SNAPBACK}>

[Rob]

Slocum's book addresses these topics in great detail:

http://www.prenhall.com/slocum/

but it seems like you might already have found your answer.

good luck and let us see how it turned out,

rob

[Martin Gamache]

Peter,

If I may revise my answer after sleeping on this. Setting the classes to an equal standard size may actually be one way to explore the data...especially in the case where your sample size or population size is the same, this is a valid way to explore for or even show relationships. As a researcher in a data exploration/visualising stage you may wish to use such an approach. As a map designer for an atlas I would stick with my earlier advice and use ranges that are more appropriate for the data.
Choosing class intervals is as much a science as an art and there are different approaches to serve different purposes. Rob's suggestion of T. Slocum's text as a reference is an excellent one.

Thanks Martin!
Peter

Set your class breaks based on the data range/distribution for each map. Although there may be relationships between factors shown on the maps, using the same class breaks for them is not the way to explore them.

<{POST_SNAPBACK}>

<{POST_SNAPBACK}>

[Pyzdra]

Once again - thanks Martin. We have been in fact doing these kinds of exploring looking at different variables inside set classes, but this was just one of the ways. For the map making purposes we will most likely follow good carto rules and reflect individual distributions of each variable.
P

Peter,

If I may revise my answer after sleeping on this.  Setting the classes to an equal standard size may actually be one way to explore the data...especially in the case where your sample size or population size is the same, this is a valid way to explore for or even show relationships. As a researcher in a data exploration/visualising stage you may wish to use such an approach. As a map designer for an atlas I would stick with my earlier advice and use ranges that are more appropriate for the data.
Choosing class intervals is as much a science as an art and there are different approaches to serve different purposes. Rob's suggestion of T. Slocum's text  as a reference is an excellent one. 

Thanks Martin!
Peter

Set your class breaks based on the data range/distribution for each map. Although there may be relationships between factors shown on the maps, using the same class breaks for them is not the way to explore them.

<{POST_SNAPBACK}>

<{POST_SNAPBACK}>

<{POST_SNAPBACK}>

[ivan]

Hi,

In this paper:

Brewer, C. A., A. M. MacEachren, L. W. Pickle, and D. Herrmann. 1997. Mapping mortality: Evaluating color schemes for choropleth maps. Annals of the Association of American Geographers 87:411-438.

The autor testing different set classes for reflect best "understanding visual" of maps.


great page



Ivan

my english is very bad ..sorry

[Pyzdra]

Thank you Ivan!
Peter

Hi,

In this paper:

Brewer, C. A., A. M. MacEachren, L. W. Pickle, and D. Herrmann. 1997. Mapping mortality: Evaluating color schemes for choropleth maps. Annals of the Association of American Geographers 87:411-438.

The autor testing different set classes for reflect best "understanding visual"  of maps.


great page



Ivan

my english is very bad ..sorry

<{POST_SNAPBACK}>