0

I have a continuous raster that has been divided in 8 zones, and I'd like to compare if the distribution of pixel values in each zone is significantly different from each other. I thought I would use a Kolmogorov-Smirnov test, but I've been advised not to.

Is there a statistical method out there that allow for the multiple comparison of non-normal distributions, without having to create a statistical model of the type Y~X?

1

I am not sure why you have been advised to not use the Kolmogorov-Smirnov test. If you put it in the context of a cumulative distribution, and not probability density, function it would likely address your question. Another option are the family of nonparametric t-tests, such as the Mann-Whitney-Wilcoxon test.

You can apply this test using the wilcox.test function in R however, the assumption of the statistic is that the samples come from a distinct population and do not affect each other. Violation of this assumption would be highly dependent on your zones.

  • Thanks Jeffrey. This main assumption seems by essence impossible to meet for spatial data...although the zones have been defined based on their hydrologic behavior and so "hydrologically" disconnected. Therefore, I think they may be considered as independent. – MerlinLutin Aug 8 '17 at 19:04

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.