6

I'm working with a time series of choropleth maps showing the same indicator for the same administrative boundaries, but spanning period of nine years.

What would be the most appropriate strategy of designing class breaks for such a time series?

At the moment, I think the quintiles division would serve the purpose the best, since it allows to scan quickly for 20% of areas with highest/lowest values (for instance).

That works for one map, however I''m not sure how to porperly choose a scheme for time series.

Options that I came up with so far would be:

  1. use class intervals created separately for each map - advantage being they reflect the data structure for each year; disatvantage - harder to compare the maps between years.
  2. use class intervals of quintiles from the summary map for the whole time period (without yearly split of the underlying data) - advantage being easier comparison of the maps for each year; disadvantage - breaks might not be ideal in certain years.
  3. use class intervals of quintiles taken from the whole distribution of the data - all areas+all years - advantages disadvantages will most likely be simmilar to the above scenario.

What are your thoughts?

3

If the purpose is to compare change over time, I think you must use the same class intervals (quintiles or otherwise) on each map, even if the breaks aren't ideal in certain years.

Including text that indicates each map is part of a series, e.g. "Map 3 in a series of 9", should cue the viewer to look for other maps in situations where they come across only one of the maps. Hopefully that will help the viewer understand why the colors look "funny" on a given map.

  • Thanks for your input. I'm also more eager to use common scales. Problem now is to which year stick to when deciding about minimum, maximum and breaks. – radek Apr 26 '13 at 13:30
  • 1
    What type of breaks do you get if you take the min/max of all data, regardless of year? I would try to work with those, unless they skew too many individual years. If so, I would pick the year that has a min and max (and maybe average) closest to the average min and average max (and maybe average) of all data. – user3461 Apr 26 '13 at 14:24

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.