On Sun, 20 Feb 2005, Volker Bahn wrote: > Dear list members, > > I created 110 spatial regression models (conditional autoregressive > regressions = CAR) in Splus (function slm()) and wanted to validated > them on reserved data. However, there isn't a method implemented with > this function to predict onto new data. Thinking about this problem I > realized that predicting onto new data would probably not be trivial for > such a model because the neighborhood matrix included in the model does > not contain the new locations. This doesn't seem to be a trivial problem > and I was wondering if people on this list are familiar with this > problem and/or solutions or work arounds. It seems that the > autoregressive models (CAR, SAR etc) would be of very limited use if > predicting onto new data was impossible.
There is a literature about this problem, which I review in: Bivand, R. S. 2002 Spatial econometrics functions in R: Classes and methods. Journal of Geographical Systems, 4, 405--421. which does look at this (for SAR). The paper has the references, in particular to Martin's work. There is, I think, no available software for creating the rectangular weights matrices to relate the included (model fit locations) to the newdata locations. It has been raised as something which should be added to the R spdep package, and is feasible with one extra layer of references (the neighbours in weights use the in-set IDs, and in this case need to allow neigbours out-of-set), I think Stéphane Dray had made some progress on this, and there is interest in this, if only to see how SAR and CAR behave on test data (in the paper, I suggest that SAR may over-fit). Please contact me off-list if you'd like to follow this up and could help, at least with real cases needing analysis. Roger > > Thank you > > Volker Bahn > ______________________________________________________________ > Verschicken Sie romantische, coole und witzige Bilder per SMS! > Jetzt bei WEB.DE FreeMail: http://f.web.de/?mc=021193 > > > -- Roger Bivand Economic Geography Section, Department of Economics, Norwegian School of Economics and Business Administration, Breiviksveien 40, N-5045 Bergen, Norway. voice: +47 55 95 93 55; fax +47 55 95 93 93 e-mail: [EMAIL PROTECTED]
* By using the ai-geostats mailing list you agree to follow its rules ( see http://www.ai-geostats.org/help_ai-geostats.htm ) * To unsubscribe to ai-geostats, send the following in the subject or in the body (plain text format) of an email message to [EMAIL PROTECTED] Signoff ai-geostats
