Modeling autocorrelation

We have so far learned how to check for spatial autocorrelation, but we have not yet learned how to incorporate this into our model. We'll now learn how to do this.

Spatial autoregression

The spatial autoregression model considers the dependence of the value upon near regions and integrates that dependence into the data-generation process. We'll be working with a Simultaneous Autoregressive (SAR) model here. One important parameter here is lambda, which indicates the level of spatial dependence, where a positive value indicates positive correlation, a negative value indicates negative correlation, and zero means no spatial dependence.

First, we fit a SAR model without a predictor (that is...