I need to perform some tests and verifications on databases and/or datasets. I need for example to analyze two databases and then determine if there is enough correlation between both.
My main issue is how to deal with correlation...
Do I need to dice/grid the map and analyze precisely each square?
Is it possible to compute statistical techniques (i.e. mean, median...) to achieve this goal or not at all? My datasets are essentially made of GeoTIFF files for elevation terrain information and Shape files for my objects (roads, trees...).

I had a look on some APIs to work with those kind of files:

Finally I read in some research papers about softwares determining databases correlation, but couldn't find any of them:

Thanks for help.


If you can avoid it, don't go for external solutions, since postgresql comes with its own statistical functions. Therefore If you can avoid it, it would be easier for you in the long run.

Now, there are many ways to correlate stuff, one good and acceptable solution is to create a dataset of random points in your AOI and add for those points the values of the coefficients you want to correlate.

Afterwards you can either correlate them using the postgresql functions or use specialized software like R, or JMP or other.

The hard part of this method is that postgis doesn't have a CreateRandomPoints function and you'll have to create one on your own. You can check here, here and here how to do so.

| improve this answer | |
  • I do not use Postgresql, I just have DEM files (GeoTIFF), ESRI shapefiles (.shp) and some imagery. These are my raw data. And eventually I can convert all of them into a VBS2 map. – eouti Apr 4 '13 at 9:47
  • Haha! yes you said databases and I assumed postgis. The underlying plan still holds: Create a random sample of points for your AOI. Then add new columns with the data you want to correlate between for those points. Afterwards you can use specialized software to correlate between values. – nickves Apr 4 '13 at 10:00
  • If you give a more detailed example of what are you trying to correlate with what we might be able to help you more – nickves Apr 4 '13 at 10:02
  • Ok, let's assume I have two (geo)databases made of data elevation model (GeoTIFF) and shapefiles. I want to assert that these 2 have same hills and plains, same roads, same urban areas... – eouti Apr 4 '13 at 10:27
  • 1
    Let's say you want to correlate two rasters. They might show the same thing, they might not. When we say X is correlated with Y we mean that a Z% of the X's values can be explained by Y's values. Same procedure, create a sample dataset of random points distributed evenly in the joint areas, load for each point the values of each dataset, and then perform the correlation analysis. – nickves Apr 4 '13 at 12:21

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.