[R-sig-Geo] SPGWR Application to Holdout Sample

[Roger Bivand]

On Tue, 29 Apr 2014, Paul Bidanset wrote:

> Hello,
>
> I was just curious - am I safe in assuming that when applying holdout data
> (new data points not included in model calibration) to GWR models in the
> SPGWR package (as below), each new point is applied to the local regression
> model in closest geographic proximity?

No, you are not safe. The #fit.points models fitted are fitted at - hence 
the name - the fit points, using the kernel and bandwidths specified. In 
this case, the data in the data= argument are geographically weighted in 
these #fit.points models, but never fitted themselves.

Please do not use HTML, and do provide an example in which georgiaNewData 
is made explicit (is it a subset of gSRDF?).

Hope this clarifies,

Roger


>
>>>> * PredictionsOfNewData  <- gwr(PctBach ~ TotPop90 + PctRural + PctEld +
> *>>>* PctFB
> *>>>* + PctPov + PctBlack, data=gSRDF, adapt=TRUE, gweight=gwr.Gauss, method =
> *>>>* "aic",  bandwidth=bw1,
> *>>>* predictions=TRUE, fit.points=georgiaNewData)
> *>>>* PredictionsOfNewData*
>
>
> Thank you in advance for your assistance.
>
> Best,
>
> Paul
>
> 	[[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
>

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 91 00
e-mail: Roger.Bivand at nhh.no