7
$\begingroup$

This question is about data visualization and statistical graphics. I have been trying to present statistical data in map. The data is at county level in the US and also at time state level. My data looks some what similar to the one below (sourced from here).

There are two issues with Choropleth maps, small counties tend to be lost and in addition small states (Like Rhode Island, Delaware) cannot be shown. My second problem is it has been hard to engage the audience in the choropleth maps. The heat map such as the one shown below does discriminate the high low values well enough.

I have read few books on data visualization, and none of them seem to have alternative presentation to choropleth maps. So my question is are there any alternative to Choropleth maps to present statistical data?

enter image description here

$\endgroup$
5
  • 1
    $\begingroup$ In addition to looking for a better visualization of the data as they are, you might also consider different things to do to your data prior to plotting, such as smoothing and/or cluster detection, depending on your goals. $\endgroup$ – D L Dahly Feb 2 '15 at 16:32
  • 1
    $\begingroup$ Have you looked at similar questions on the Geographic Information Systems site? $\endgroup$ – whuber Feb 2 '15 at 16:50
  • 1
    $\begingroup$ Can you elaborate on the second problem? I've thought most viewers find maps to be more engaging. An additional problem is that big counties give a false impression of uniformity, though that can be solved with a geospatial scatter plot. $\endgroup$ – xan Feb 2 '15 at 18:03
  • 1
    $\begingroup$ @whuber, I did not know that there was GIS site, thanks for pointing it out, I'll definitely check it out. $\endgroup$ – forecaster Feb 2 '15 at 19:04
  • $\begingroup$ @xan, I annot get the audience attention to focus on high/medium counties in the above example and especially small counties in LA, MS and AR. In addition counties in MD/DE/RI. I was thinking of discretize and then plot several different maps. $\endgroup$ – forecaster Feb 2 '15 at 19:09
5
$\begingroup$

Choropleths have a number of flaws, as you note. Most infamous is the way a shape size is usually unrelated to it's measure yet the size is very prominent visually (electoral maps are a classic example).

Cartograms strive to solve the sizing issue but distort the geography, which looks odd and can be a problem if you're looking for geographic patterns.

A few common alternatives more applicable to US counties:

Geographic Scatterplot

Draw a dot for each shape. That way every shape gets the same amount of color, though overstriking is still an issue for very tiny shapes.

enter image description here

Micromaps

Works with choropleths or geographic scatterplots. Partition the graphs by geographic sector or some measure, not necessarily the same as the coloring variable.

enter image description here

Custom Coloring

You mentioned needing to focus on a particular range of values. One way to help with that is to use a coloring scheme that highlights that range at the expense of others.

enter image description here

Smooth Contours

It sounds like you want to see individual shapes, but if you're looking for broad patterns, plotting a smoothed contour like in a weather map can be useful (no picture).

No Map

Finally, if the data values are more important than the geographic patterns, consider another kind of graph altogether, such as a ranked bar chart of the top and bottom counties. The general weakness of maps is that they use the two most prominent dimensions (X and Y) for geography and leave lesser dimensions of the data measures, so the geography must be relevant to justify using a map.

$\endgroup$
1
  • $\begingroup$ +1 Thanks for for your effort in provding a detailed answer. I just ordered the book by Carr and Pickle on Linked Micromaps which looks like a promising technique. $\endgroup$ – forecaster Feb 4 '15 at 16:43
5
$\begingroup$

Here is a slew of examples (hopefully that's OK) which try to show variations on the map theme while providing a range of flexibility that is lacking in the standard choropleth. This may overlap other answers, but I am trying to be as exhaustive as possible.

Favorite: Interactive density with multiple levels of scale

Before going through them, I will say that the "racial dot plot" of US census data is one of the most compelling visualizations that solves this problem. It is highly interactive, handles scale, and allows for an incredible density of information. Creating it may be difficult, but it sure is beautiful. Check it out first: http://demographics.coopercenter.org/DotMap/index.html

Racial dot map

After that, here is a more systematic take on the examples.

Option 1: plot the data based on the size of an object instead of coloring an area. This can be with a bubble, density, or weather map style.

Option 2: extract the shapes of the counties/states and plot them small multiple fashion with the same scaling

Option 3: allow for a degree of interactivity so that the scale can be changed at will by the user

Option 4: have multiple visualizations that highlight areas of interest at the small scale

Large scale of baseball Small scale baseball

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.