[R-sig-Geo] a comparison question

[Hodgess, Erin]

Thanks to both very much!
________________________________________
From: Roger Bivand [Roger.Bivand at nhh.no]
Sent: Friday, March 07, 2014 10:35 AM
To: Wissner, Michael (DOF)
Cc: Hodgess, Erin; r-sig-geo at r-project.org
Subject: Re: [R-sig-Geo] a comparison question

On Fri, 7 Mar 2014, Wissner, Michael (DOF) wrote:

> Erin,
>
> My (basic) understanding of these two techniques is that kriging outputs
> a predictive surface with values derived from a variogram while GWR
> executes a regression equation on each feature in a dataset with a
> spatial weighting component.  So I think the answer to your question is
> that it depends on what output you're looking for and what you're trying
> to accomplish.
>
>> From ESRI's How Kriging Works
>> (http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009z00000076000000.htm)...
>> "it has been said that kriging uses the data twice: the first time to
>> estimate the spatial autocorrelation of the data and the second to make
>> the predictions."
>
>> From ESRI's How GWR Works
>> (http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//005p00000031000000)...
>> "GWR provides a local model of the variable or process you are trying
>> to understand/predict by fitting a regression equation to every feature
>> in the dataset. GWR constructs these separate equations by
>> incorporating the dependent and explanatory variables of features
>> falling within the bandwidth of each target feature."
>
> Thus, kriging fits a model based on the paired distances then applies
> that to predict an entire surface of a single z-value (variable). I
> think consensus among geostatisticians says that kriging is the best for
> spatial prediction of a single variable. GWR permits a whole set of
> explanatory variables to be analyzed and included in a spatially
> weighted regression model unique to each feature.  I think this
> technique more suited for exploring how a complete set of independent
> variables varies in its explanatory power across the spaces analyzed.

OK with regard to kriging; ESRI's help page for GWR does not refer to:

Paez A, Farber S, Wheeler D, 2011, "A simulation-based study of
geographically weighted regression as a method for investigating spatially
varying relationships", Environment and Planning A 43(12) 2992-3010

which spgwr::gwr does refer to. Until you have read and understood this
article, never even consider using GWR, irrespective of what ESRI or the
authors of GWR say. Not only are there local collinearity (mentioned by
ESRI) problems, but GWR creates patterns in local for random input data,
see the gwr() examples. Use, if at all, very carefully.

Hope this helps,

Roger

>
> I hope this is useful.  I am by no means an authority on this and will
> keep a close eye on this thread.
>
> Chao, mike
>
>
>
>
> -----Original Message-----
> From: r-sig-geo-bounces at r-project.org [mailto:r-sig-geo-bounces at r-project.org] On Behalf Of Hodgess, Erin
> Sent: Wednesday, March 05, 2014 3:03 PM
> To: r-sig-geo at r-project.org
> Subject: [R-sig-Geo] a comparison question
>
> Hello everyone:
>
> Here is a comparison question, please:  is kriging better than GWR, please?  Or are they even comparable, please?
>
> Thanks,
> Erin
>
>
>       [[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
>
> _______________________________________________
> 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 95 43
e-mail: Roger.Bivand at nhh.no