[R-sig-Geo] Predict

[Roger Bivand]

On Fri, 30 Mar 2012, Smit, M.J. wrote:

> Dear list members,
>
> I want to do a scenario analysis: run a regression on the real data, then make some changes to that, and do a prediction based on the changed data. There is a predict command under sarlm for this, and it works. However, I have two questions.
>
> - predict(model,newdata=NULL,weights) uses not only trend (the 
> non-spatial terms) and signal (the spatial "smooth") but also noise (the 
> residuals from the original regression). Is it true I can avoid this by 
> explicitly inserting my old dataset into newdata=? The predictions 
> differ, so something has happened.

Yes, this is what is happening. In the newdata=NULL case, the response is 
known, so the signal is defined in terms of the response. If newdata is 
given, it is defined in terms of the data generation process.

> - predict then gives me a list object, and I'm at a loss how to get the 
> results from this. I've named the objects pred1 and pred2, and vainly 
> tried pred1$trend and pred1[[1]], which gives the first observation from 
> the $trend subvariable, but doesn't allow access to the other 
> subvariables. Is there a way to get this into as.data.frame?

I'll add an S3 method to do this. For now, assign the result of the print 
method to an object:

goAway <- capture.output(predsDataFrame <- print(<sarlm.predObject>))

with capture.output() to prevent it printing to screen. The 
as.data.frame() method is in revision 430 on R-forge.

Roger

>
> Best regards,
> Martijn
>
> _______________________________________________
> 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, 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