[R-sig-Geo] posting on behalf of Kevin, Yen

[Roger Bivand]

I'm posting this on behalf Kevin, Yen - the same text was trapped 
yesterday as spam, so this is also a technical issue, and the reason for 
helping in this way:
-----------------------
Hi, lists,

I am reading meterials about GWR, but there are some problems puzzling me 
very much:

If there are N observations at N locations g(g1,¡­¡­gn),the dependent and 
independent variables are Y(y1,...yn) and X(x1,...xn). The formula for the 
regression is Y=X*Beta, where Beta is the function of location g.

My questions are as follows:
   1.How many groups should be grouped for regressing? N groups for the N 
locations or M groups (M is a arbitrary number less than N)?
   2.When the parameter beta(gi) at location gi is estimated, weight Wi is 
used. So how is W used? Are the neighboring k(with in the bandwidth h) 
observations used for calculation?
   3.Beta is the function of location. Is Beta same as W or expressed as a 
function including W like Beta=W*f(g)?

Many thanks and best wishes

Kevin,Yen
LIESMARS,Wuhan University,China

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, 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