RE: st: Spatial correlation?

["Nick Cox" <n.j.cox@durham.ac.uk>]

Much depends on the details of the problem, including whether 

points are paired or not 

there is some background of theory or experience that indicates what precise comparisons make most scientific or practical sense

how distance between points is best defined, or even whether similarity of locations is best assessed as a distance any way. 

If points were paired and distance between them can be easily measured, then looking at the pattern of pairwise displacements (distances and directions) would seem natural. 

A reference has sprung out of some dark recess in my memory

Tobler, W.R. 1965 Computation of the Correspondence of Geographical Patterns. 
Papers, Regional Science Association, 15: 131-139. 

(I didn't remember volume number and pages, but the internet helps.) 

But it's about 40 years since I read it and I don't recall anything beyond a clever use of correlations with complex numbers (real and imaginary parts). But there must be more here and there. 

Nick 
n.j.cox@durham.ac.uk 

Karin Jensen

Thanks Nick. Is there a more complex method that you would recommend?

On 21 June 2010 09:38, Nick Cox <n.j.cox@durham.ac.uk> wrote:

> Thanks to Stas for the plug, but -variogram- (SSC) does not include direct support for this kind of analysis. I don't think there is a simple, single canned method for what you want.

Stas Kolenikov

> On Sun, Jun 20, 2010 at 9:09 AM, K Jensen <k.x.jensen@googlemail.com> wrote:
>> Is there a way of calculating an equivalent to a correlation
>> coefficient for geographical data in Stata?  I have latitude and
>> longitude for two types of data point and would like to measure the
>> strength of the association between the two.  Both types have
>> autocorrelation (points are likely to be near each other within the
>> same dataset).
>
> findit variogram

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/