 |
 |
Hello all I'm interested in testing for differences between my fixed sample (n=200) representing locations of colonies (of a vole species) and N random samples using permutation test. The data are provided by pixel's reflectance of satellite images. My fixed sample are georeferenced locations in the field and I'm obtaning N random samples (same size, n=200) from the entire image. Can I do this? obtain random sub-samples from an known finite sample (the entire satellite image, with about 1,000,000 pixels) and compare N times with one fixed sample (the colonies reflectance), using an nonparametric statistic? What about sampling, say, Nc from this pool of 200 colonies points and another Ns from satellite (Nc=Ns) and run the statistical test with this pair of sub-samples? by taking sub-samples from each one, is a valid approach? I implemented an R function running a large number of wilcox test, evaluating distribution of p-values and using median of p-values as a location to test my null hipothesis. Now I'm worling in implement bi-aspect nonparametric permutation test, since random samples from satellite can show large sd. any idea?