These statistical tests will only give you an idea on how these two matrixes
differ but they do not take into account connectivity between cells, and
dates of samples. And it seems that the difference interests you.
Another test you could use is testing for spatial autocorrelation on using
spdep package, after having established a list of neighbours for each cell
with cell2nb.
The test for sp. autocorrelation can either be global either local.
It seems that you have some space time data. Maybe you should have a look at
spacetime package, also?
I'm just beginning in spatial statistics. Someone will tell me if I'm wrong.
hth,
Mathieu
2011/9/1 Mathieu Rajerison <mathieu.rajerison at gmail.com>
Giulia,
You may look at Pearson, Student or Kolmogorov Smirnov test
cor.test, t.test, and ks.test
after having coerced your matrixes into vectors
Maybe someone has another idea or thinks my advice is not right?
hth
Mathieu
2011/9/1 Conchedda, Giulia (AGAL) <Giulia.Conchedda at fao.org>
Hi to All,
My question concerns the best practice to compare cell-by-cell two grids
that are representing the same phenomenon (distribution of rural poor). The
grids have the same spatial extent and cell size and I would like to
quantify and provide statistics on how pixel values in the two grids differ.
I first subtracted one from the other and in this way I can already say and
spot where differences are more frequent. What is the best statistics to
quantify this difference?
Thanks for advising,
Giulia