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  
morglum

Registered:
Posts: 2
Reply with quote  #1 
Hi everyone
 
I have read and very enjoyed "Show me the numbers".   I was wondering if you have some advice on visualising transition matrices, such as this matrix on interprovincial migration.
  
The circular plots such as this one are growing increasingly popular, but I'm not sure I enjoy having to follow a line through multiple lines to figure out where people actually come from and go to.
 
Could this be a good case of "the table you have used for  the last 100 years is just fine?"

Maybe an interactive map showing only the arrows to/from a single state could be useful ?
Thanks,
Simon
sfew

Moderator
Registered:
Posts: 827
Reply with quote  #2 
Hi Simon,

Tabular displays with numbers, such as the first example that you linked to, work perfectly if the primary purpose is to look up individual values. For this purpose, no benefit is gained by display the data graphically. If rapid comparisons between the values are needed among a relatively small data sets, such as the one about interprovincial migration, the same tabular arrangement can be used with horizontal bars rater than numbers. With larger data sets, however, circular displays such as the one that you linked to are often the more effective form of display.

__________________
Stephen Few
morglum

Registered:
Posts: 2
Reply with quote  #3 
Hi Stephen,
Thanks for replying!

I built a mock-up to make sure I understand what you mean by "tabular arrangement with horizontal bars".    Is this what you mean?  I thought I'd keep the numbers since otherwise the small provinces have no blue visible.

unnamed.png 
It is missing a table title, and something to explain that the receiving province is the "column" and the sending province is the "row".  I'll probably get rid of the horizontal borders when I learn how to do it too.  Thousand separators will also be added once I learn how to do it.  


sfew

Moderator
Registered:
Posts: 827
Reply with quote  #4 
Simon,

Unfortunately, after playing with the data a bit myself, I found that the tabular arrangement of horizontal bars doesn't work well with this data set because of the huge scaling differences among the values, from a low of 8 to a high of 27,648. This scaling difference causes most of the bars to be so minor in length that they're of little use.

If this scaling problem did not exist, your design would be improved by having the bars from from left to right rather than right to left, removing the numbers, and adding vertical lines to show the left and right edges to each cell.

With this data set, the circular chart probably works best if a grahical display of the data is actually needed.

__________________
Stephen Few
heinzel

Registered:
Posts: 16
Reply with quote  #5 
If completeness of the data is not a key requirement, one could list the top 10 relationships in a one-dimensional table or bar chart and mention that all other relationships are smaller than X.
danz

Registered:
Posts: 190
Reply with quote  #6 
I agree with heinzel that a sorted bar chart of top relationships are a good addition to the matrix.

One extra issue for circular plots with many relationships appears when data is bidirectional, having asymmetric values (A->B != B->A). Double density arcs are required, orientation marked by arrows, making the whole graph even more difficult to read.

In case the values have an additive sense and they are asymmetric (as in presented case), a dual stacked bars solution looks to me a better choice to circular plots.
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.