Discussion


Note: Images may be inserted into your messages by uploading a file attachment (see "Manage Attachments"). Even though it doesn't appear when previewed, the image will appear at the end of your message once it is posted.
Register Latest Topics
 
 
 


Reply
  Author   Comment  
bpierce

Moderator
Registered:
Posts: 100
Reply with quote  #1 

For the October/November/December 2016 Visual Business Intelligence Newsletter article, titled The Visual Perception of Variation in Data Displays, Stephen illustrates and explains the scientifically based best practices for the most common—and perhaps the most important—of data sensemaking tasks: examining variation.

What are your thoughts about the article? We invite you to post your comments here.

-Bryan

danz

Registered:
Posts: 190
Reply with quote  #2 
In order to improve the visual decoding eficiency I have some remarks.

Limit the amount of encoding techniques per view
Scatter charts are extremely powerful representations. In theory we can use X, Y, Color (categorical and sequential palette), Shape Type, Shape Size to encode different variables, numerical and categorical. We can use convex or concave hulls (polygons) to enclose certain values for grouping purposes as well. Within the same chart we can use a categorical color palette for regions each color with limited length sequences of 2-3 intensities for encoding consecutive years, most intense for current year. The fact we can encode that much in one view, doesn't mean we need to use all known techniques at once. I always suggest a limit of 3 (exceptionaly 4) encoding techniques per visualization.

Colors used for quantitative encoding
Color intensity can be used in some limits for quantitative encoding. I am very much aware of the beauty of a gradient color scale provided by most vendors to encode quantitative values within some limits. How efficient is a continuous color scale? I consider that a sequential or divergent palette of distinct colors associated with intervals will always be the better choice.

Multiple graphs instead of one
Four scatter graphs with the same scales next to each other might be a better alternative to a dense one where regions are color or shape encoded. Multiple  choropleth maps of the same map next to each other are sometimes a better choice to one heavy colored map.

Auxiliary constructions
When amount of data we need to display/analyses is large, usually window techniques are used. However, for global picture, it is always an idea to use some statistical constructions related to the analyses we need to perform. Average reference lines, confidence and prediction bands, regression curves, Tukey bagplots are just a few.

Dan
Berry

Registered:
Posts: 6
Reply with quote  #3 
Impressive article, thanks for the overview of attributes!

On not using hue for quantitative data (page 6/7):
I find that colorramps work quite well for geographical information even if not displayed on a map, see e.g.

[bathymetry] 
Data generating and processing code along with spatially interpolated version on
https://github.com/brry/misc/tree/master/bathymetry

What do you think of that from the perception point of view?

(I intentionally used a color ramp that also scales in lightness. Using only lightness appears to me less appealing than also using hue, see black/white print below. More on color e.g. here: http://vanseodesign.com/web-design/hue-saturation-and-lightness/)
bathymetry_bw.png

wd

Registered:
Posts: 167
Reply with quote  #4 
Re: "We can only display quantitative variation effectively if we understand visual perception."

Agree, but I would add that we also need to know how to do data analysis. Once we understand the story that's in the data. the visual story telling isn't so difficult.

__________________
Bill Droogendyk
sfew

Moderator
Registered:
Posts: 837
Reply with quote  #5 
Berry,

I see that by "color ramp" you mean variation in hue. It is true that our ability to distinguish variations in color intensity (i.e., in lightness, saturation, or both working in concert) can be enhanced using slight hue variation as well. This only works, however, if color intensity variation is present, otherwise the hue variation would represent an intuitive sequence. Color experts such as Cynthia Brewer and Maureen Stone often add slight hue variations in sequential color scales the enhance our ability to discern quantitative variation. To do this effectively, as Cindy and Maureen do, you must understand color well enough to select hues that work well together for this purpose. The example that you included, which varies from yellow through green and blue to end with indigo, works well. In fact, it looks like one of Cindy Brewer's sequential palettes.

__________________
Stephen Few
sfew

Moderator
Registered:
Posts: 837
Reply with quote  #6 
Bill,

There are many things that we must understand other than visual perception to visualize data effectively. Basic skills in data analysis are definitely required, as I often point out in my work. Skills in data analysis, however, are built on several areas of understanding, including visual perception.

__________________
Stephen Few
wm85

Registered:
Posts: 1
Reply with quote  #7 
I was surprised that the only examples given for displaying spatial variation were bubble plots on a map and a choropleth map.  A recent issue of the Journal of Statistical Software had several articles on geovisualization, including two on linked micromaps.  As described in the articles linked micromaps have several advantages over choropleth maps.  Here is the link to that issue:  https://www.jstatsoft.org/issue/view/v063.

sfew

Moderator
Registered:
Posts: 837
Reply with quote  #8 

wm85,

All of the "linked micro-maps" that appear in the article that you mentioned are either choropleth maps and bubble maps. A micro-mao is a series of small maps that are meant for comparison. A micro-map is an example of what Tufte called "small multiples" back in 1983 and what other called trellis graphs for even longer. Regardless of what we choose to call them, they are not a new graph type, but merely a collection of graphs that vary along a particular dimension, such as time. They are extremely useful and I have written about them extensively.


__________________
Stephen Few
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.