[R-sig-Geo] Degrees of freedom in GWR

[Roger Bivand]

On Tue, 9 Jul 2013, Michael Blows wrote:

> Hi all, can any body clarify me why the degrees of freedom in R output 
> of ?GWR comes in decimal? Mike

Please read the references, where this is explained - why would you expect 
them to be integer when weighted proportions of observations are being 
used for each fit point? The print method for fitted gwr objects reports 
two versions of effective degrees of freedom if the hat matrix is 
computed, one complete, the other an approximation used in the 2002 GWR 
book, but these non-integer numbers are also based on assumptions.

> 	[[alternative HTML version deleted]]

Please only post plain text!

Roger

>
>

-- 
Roger Bivand
Department of Economics, NHH Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no