5

I'm working on a bubble chart, and am having trouble calculating the bubble radius.

I want to use a certain quantity to calculate the bubble radius, but the quantity varies a lot. The minimum is 340, maximum is 190000 but most points are between 5000 to 30000. These ranges are too extreme to use the quantity as the radius directly.

Is there any algorithm to "smooth" the curve of the data? Or some method to map the original quantity to a more reasonable one?

Right now my graph is a scatter chart, because I can not translate it to bubble chart without a radius.

phs
  • 10,687
  • 4
  • 58
  • 84
Jim
  • 51
  • 3
  • If you map the values to some other range, how would you make sure that they still adequately represent the quantities? Do you want to preserve the "clustering" of values in the image of [5000, 30000]? I'm not sure a bubble chart will be the best option here; maybe a scatter chart with a color scale for the quantity will be more appropriate? – G. Bach Apr 25 '15 at 02:03
  • log(value) is routinely used to "linearize" extreme values and reduce the range. – Mörre Apr 25 '15 at 02:04
  • Because people cann't recognize magnititude very good , I just use the bubble to approximately represent the quantities.There is no too much accurency needed.The 'color' is already used to represent department infomation.log(value) is quite a smart method: ) – Jim Apr 25 '15 at 02:23

2 Answers2

2

Have you considered using the logarithm of your data? Using base 10 for example, your bounds become

| Original | Log_10 |
|     340  |    2.5 |
|   5,000  |    3.7 |
|  30,000  |    4.5 |
| 190,000  |    5.3 |
phs
  • 10,687
  • 4
  • 58
  • 84
0

Something else you might consider, if you haven't done so already, is taking the square roots of your data and using those as the radii, so the area of the bubbles are proportional to the numbers. Then your radii will vary from about 20 to about 500, a ratio of roughly 20:1.

Edward Doolittle
  • 4,002
  • 2
  • 14
  • 27