[R-sig-Geo] Θέμα: Moran I and General Trend

[Sotos]

Many thanks for you reply, 

But wouldn't Local moran be better for checking for
autocorellation and local clusters?

Also I read that in SIDS dataset, Freeman-Tuckey
square root transformation was used because the
variance was not constant. Would this be appropriate
to use for crime rates with non constant mean? I guess
that a lot depends on the actual data, but are methods
appropriate or am I missing something?

Cheers,
S.



--- Nicholas Lewin-Koh <nikko at hailmail.net> έγραψε:

> Hi
> If you have a first order effect like that then you
> can 
> try fitting a linear model to the data and the use
> the residuals for checking autocorrelation. This
> will
> introduce some bias but it is small relative to the
> gain if
> the residuals are stationary. 
> 
> Nicholas
> 




	

	
		
___________________________________________________________ 


διαθέτει την καλύτερη δυνατή προστασία κατά των ενοχλητικών 
μηνυμάτων http://login.yahoo.com/config/mail?.intl=gr