I'm working on a project to estimate real estate and started with some classique techniques, such as linear regression etc. The obtained results are already going in the good direction, but to get some more precise results, I've started to lookup some more advanced models.

First I've looked into a spatial autoregressive model, but these models are a bit hard to use if we want to predict new values, or am I wrong?

The next one is the kriging technique. In general, these models use only two coordinates and one prediction value (like temperature or height). In my project I'm using way more variables to predict the outcome.

My question is which technique/model I would best look into to obtain some better results.


I prefer autoregressive models to kriging, because I feel there is more of an economic rationale behind their use. Others are free to disagree!

Autoregressive models may be difficult to use/understand/apply, but if there is spatial correlation in your data (and there almost certainly is in home price data), linear regression models are biased and hypothesis tests on them are invalid. In fact, you will probably get less precise estimates, but it is better to be accurate and imprecise than highly precise but off the mark.

I highly recommend the spdep package for R by Roger Bivand. It's relatively easy and quick to use, and it intersects well with the nomenclature used by LeSage and Pace in their excellent book.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.