Issue with ColorFunction in GeoRegionValuePlot?

Questions
, , ,
1

I have a list of towns with numbers assigned (population changes). I want to colour the towns green or red according to whether their population changes are positive or negative. My thought was I would use a custom ColorFunction as follows:

ColorFunctionScaling -> False, ColorFunction -> (If[# > 0, Green, Red] &)

However, this does not work. All the points come out red, even though many are positive. If I set the threshold for distinguishing green from red at -5600, then it acts as though the threshold is somewhere around 0. In general, the colouring of points on the map does not reflect the threshold I use in the ColorFunction.

I have played around and reproduced the problem in simple form as follows:

GeoRegionValuePlot[{Entity[
    "City", {"London", "GreaterLondon", "UnitedKingdom"}] -> 334, 
  Entity["City", {"Paris", "IleDeFrance", "France"}] -> 356, 
  Entity["City", {"Berlin", "Berlin", "Germany"}] -> 399}, 
 ColorFunctionScaling -> False, 
 ColorFunction -> (If[# > 358, Green, Red] &)]

Clearly, London and Paris should be coloured red (their values are less than 358) while Berlin should be coloured green. However, the actual result gives Paris as green, falsely implying that its value (356) is greater than the threshold of 358.

GRVPproblem.PNG.jpg
GRVPproblem.PNG.jpg675×381 95.2 KB

2

I’m using Version 10.1 running Windows 7. Executing code in your In[161], Paris and London were red and Berlin was green.

3

Hi,

I am on a Mac and if I execute:

GeoRegionValuePlot[{Entity[
    "City", {"London", "GreaterLondon", "UnitedKingdom"}] -> 334, 
  Entity["City", {"Paris", "IleDeFrance", "France"}] -> 356, 
  Entity["City", {"Berlin", "Berlin", "Germany"}] -> 399}, 
 ColorFunctionScaling -> False, 
 ColorFunction -> (If[# > 358, Green, Red] &)]

I get

Screenshot2019-02-0222.09.56.jpg
Screenshot2019-02-0222.09.56.jpg517×298 61.4 KB

as in the OP. I get the same if I execute that in the Wolfram Cloud.

Cheers,

Marco

4

I’m using Version 11.3 on Windows 8.1.
I’ve checked with Version 11.2, and I get the same.

5

I continue to investigate this, and I don’t seem to be able to find any workaround. Nothing works as I would expect. I have tried setting ColorFunctionScaling->True, and setting my thresholds to the resulting normalised values, but this gives incomprehensible results–for example, even though all values are positive and I set the threshold at 0, they are all coloured red as if below the threshold.

Other bizarre behaviour…

Trying a different approach to defining a ColorFunction using Hue (not useful for my intended purpose but just testing the performance of this function)…

GeoRegionValuePlot[{Entity[
    "City", {"London", "GreaterLondon", "UnitedKingdom"}] -> 0.1, 
  Entity["City", {"Paris", "IleDeFrance", "France"}] -> 0.4, 
  Entity["City", {"Berlin", "Berlin", "Germany"}] -> 0.7}, 
 ColorFunctionScaling -> False, ColorFunction -> Function[u, Hue[u]]]

GRVPproblem2.PNG.jpg
GRVPproblem2.PNG.jpg481×286 57 KB

Looks feasible. I then change the value of London, which should not change the colours of the other two cities, since their values do not change, but in fact it does.

GeoRegionValuePlot[{Entity[
    "City", {"London", "GreaterLondon", "UnitedKingdom"}] -> 0.2, 
  Entity["City", {"Paris", "IleDeFrance", "France"}] -> 0.4, 
  Entity["City", {"Berlin", "Berlin", "Germany"}] -> 0.7}, 
 ColorFunctionScaling -> False, ColorFunction -> Function[u, Hue[u]]]

This gives…

GRVPproblem3.PNG.jpg
GRVPproblem3.PNG.jpg502×287 58.3 KB

I think, maybe it is acting as though ColorFunctionScaling is True (even though I set it to false). In this case, changing the value of London changes the minimum value, which affects the scaling. So I try changing the value for Paris, which is between the extremes. In this case, it looks as though the ColorFunction is behaving in some quantised way, even though it ought to be continuous.

With Paris changed to 0.3

GeoRegionValuePlot[{Entity[
    "City", {"London", "GreaterLondon", "UnitedKingdom"}] -> 0.1, 
  Entity["City", {"Paris", "IleDeFrance", "France"}] -> 0.3, 
  Entity["City", {"Berlin", "Berlin", "Germany"}] -> 0.7}, 
 ColorFunctionScaling -> False, ColorFunction -> Function[u, Hue[u]]]

I get the following, where Paris is now coloured the same as London…

GRVPproblem4.PNG.jpg
GRVPproblem4.PNG.jpg488×287 57.4 KB

And if I change Paris to 0.5…

GeoRegionValuePlot[{Entity[
    "City", {"London", "GreaterLondon", "UnitedKingdom"}] -> 0.1, 
  Entity["City", {"Paris", "IleDeFrance", "France"}] -> 0.5, 
  Entity["City", {"Berlin", "Berlin", "Germany"}] -> 0.7}, 
 ColorFunctionScaling -> False, ColorFunction -> Function[u, Hue[u]]]

I get the following, where Paris seems to be coloured the same as when its value is 0.4 (even though Hue[0.5] ought to look different from Hue[0.4]…

GRVPproblem5.PNG.jpg
GRVPproblem5.PNG.jpg485×286 57.5 KB

I’m beginning to think that ColorFunction within GeoRegionValuePlot got broken somewhere between Mathematica 10 and Mathematica 11.3, with no logical explanation or workaround, and so it is now impossible to achieve what I want..?

6

Did test link you try doing this in an earlier version of the software, Marc?

7

I have not tried this in an earlier version.

I have found a workaround by first mapping the data onto their colours, then passing the new data to GeoRegionValuePlot. I.e., I do

data2 = Map[#[[1]] -> If[#[[2]] > 0, Green, Red] &, data]
GeoRegionValuePlot[data2]
8

I reported the bug to Support so that it will be addressed in the upcoming Version 12.

Regards,

9

Thank you very much.

10

The bug is still there in Version 12, unfortunately.

11

Thanks for sharing the workaround, Marc. I used it to ensure the correct color values in the RegionValuePlot in Mathematica version 12.

Today, I ran into an issue with a custom ColorFunction in RegionValuePlot, which showed colors for US Sates inconsistent with their corresponding values.

Cheers,

Dave

12

I am seeing a similar issue. I am using Mathematica 12.2.0.0 on Mac OS10.15.7. Is this still a bug? Is there a solution besides mapping the data as shared by Marc (above)? P.S. This is my first post so I’m not sure how this works…

bl = BarLegend[{palette, {0, 4}}, LabelStyle -> 10];

getColor[val2_] := ColorData["TemperatureMap", val2/4];

{getColor[.1], getColor[1], getColor[1.5], getColor[2], getColor[2.5],
  getColor[3], getColor[3.5], getColor[4]}

gPlot = <|Entity[
    "AdministrativeDivision", {"FairfieldCounty", "Connecticut", 
     "UnitedStates"}] -> Log10[10000.], 
  Entity["AdministrativeDivision", {"HartfordCounty", "Connecticut", 
     "UnitedStates"}] -> Log10[5000.], 
  Entity["AdministrativeDivision", {"LitchfieldCounty", "Connecticut",
      "UnitedStates"}] -> Log10[1000.], 
  Entity["AdministrativeDivision", {"MiddlesexCounty", "Connecticut", 
     "UnitedStates"}] -> Log10[500.], 
  Entity["AdministrativeDivision", {"NewHavenCounty", "Connecticut", 
     "UnitedStates"}] -> Log10[100.], 
  Entity["AdministrativeDivision", {"NewLondonCounty", "Connecticut", 
     "UnitedStates"}] -> Log10[50.], 
  Entity["AdministrativeDivision", {"TollandCounty", "Connecticut", 
     "UnitedStates"}] -> Log10[10.], 
  Entity["AdministrativeDivision", {"WindhamCounty", "Connecticut", 
     "UnitedStates"}] -> Log10[5.]|>

GeoRegionValuePlot[gPlot,
 ImageSize -> 400,
 PlotRange -> {0, 4},
 ColorFunction -> (getColor[#] &),
 ColorFunctionScaling -> False,
 PlotLegends -> bl,
 PlotStyle -> Directive[Opacity[.6]]]

Image.jpg
Image.jpg1558×1114 146 KB

13

Can confirm this issue again in Mathematica version 12.2 on Windows 10 build 19042.844 64-bit.

Marc’s work-around still works. Needed to create a GeoRegionValuePlot to update an analysis with new data and ran into this and a few more issues.

Unfortunately, the situation seems to have gotten a bit worse: using your own ColorFunction no longer works (it used to show some incorrect results). Replacing the legend using this method /._Barlegend->Legend also no longer seems to work.

To my surprise I no longer seem to be able to create a plot overlay with Show[{GeoRegionValuePlot ,GeoListPlot,GeoGraphics}].

This generates e.g.

  • Rule::argr: Rule called with 1 argument; 2 arguments are expected.
  • Could not combine the graphics objects in Show

I will look into it a bit further, but the individual Geo plots can be created successfully, but when combined in Show, Mathematica throws an error.