[R-sig-Geo] A question about the estimated standard errors in SPGWR package

Danlin Yu yud at mail.montclair.edu
Tue Jul 14 18:32:22 CEST 2015


Hi, I would assume x_se_EDF is calculated with the effective degree of 
freedom, while x_se is calculated with the common sense degree of 
freedom (number of cases minus number of variables minus one). In any 
cases, the effective degree of freedom shall be used in GWR estimation 
since that is the "actual (effective)" degree of freedom in estimation.

I hope this helps.

Danlin

On 2015/7/14 23:37, Apostolos Tsimpanos wrote:
>   
>
> Dear List
>
>   
>
> I have a simple question about the estimated standard errors obtained from
> the gwr command in SPGWR package. I run a  GWR  regression y~x
>
> GWRmodel<-gwr(y~x,coords=cbind(XCOO,YCOO),bandwidth=gfixcvsel,gweight=gwr.bi
> square,hatmatrix=TRUE,se.fit=TRUE,se.fit.CCT=TRUE)
>
> In the data frame
>
> GWRmodel$SDF
>
> there are two estimated standard errors for the x coefficients denoted as
> x_se and x_se_EDF.
>
> What is the difference between x_se and x_se_EDF and what is the best for
> use, the former or the latter? The manual lacks information about this.
>
> If I am not confused I think that the difference should be in the
> approximation of hat matrix  that is  x_se  uses s^2 with approximation
> whereas x_se_EDF not.
>
> I would appreciate any help.
>
>   
>
> Thanks in advance
>
> Apostolis
>
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo



More information about the R-sig-Geo mailing list