Skip to content
Prev 851 / 7420 Next

Reducing spatial autocorrelation

My apologies to Carsten for misspelling his name, 
but I still see its paper appropriate in the multivariate case.

I was thinking on  something like computing your 
favourite ordination or NMDS on the species or 
environmental or both (i.e RDA), and then compute 
a correlogram on your ordination scores. This 
would let you select the minimum distance between 
sites to minimize autocorrelation (this could 
also be obtained from a multivariate Mantel correlogram).
And /or you could also generate autocovariates 
from your ordination score(s) and include them as 
predictors in that "some statistical analyses" to 
account for the autocorrelation.


Marcelino
At 15:41 14/10/2009, Matthew Landis wrote: