Define colors in bivariate scatterplot according to third raster layer

[pvanbosgeo]

I don't know if this is feasible, but what could greatly enhance the scatter-plot tool is the option to define the point colours according to a third raster layer. This would offer different ways to visually link spatial patterns to e.g., patterns in environmental space. Think e.g., plotting rainfall against temperature, and colouring points in the plot according to altitude layer, or plotting poverty against market access, colouring points according to political units.

[wenzeslaus]

This should be relatively easy for G7:d.correlate. How hard would be to add it to the GUI (like G7:d.histogram) is another thing. I'm not sure how about the wx-based scatter plot which is what you mean I suppose.

But since you asked, I finished r.scatterplot in r69716 which is derived from r3.scatterplot and creates the scatter plot as vector map with points, see:

https://grass.osgeo.org/grass70/manuals/addons/r.scatterplot.html

A scatter plot as vector points as opposed to an image has some pros and cons. Some of the pros for the user are: you can measure and query in the "image" using (all) the normal tools, visualize with full vector visualization tool set, further select from the data, and you can processes the data (to e.g. bin it). A significant con is (lower) speed (if nothing else, you first iterate over the data to create the vector and then once again to draw).

Here is an example result from the manual page:

This concept is basically what is in R or Matplotlib where the spatial data are often drawn using the originally non-spatial displays, here we convert the non-spatial to spatial and use the spatial data display.

[pvanbosgeo]

This is great, a very interesting approach, and it seems to do precisely what I was after. In fact, it is even better as it allows 3D graphs.

[martinl]

Still relevant?