
449Spatial Auto-Correlation and Auto-Regression
First, assume that X3 sub-population is more prone to the disease, and therefore X1 is modeled
as a function of X3
> xd.lag <- lagsarlm(X1 ~ X3, data=xd, xd.nblsw, tol.solve=1e-12)
> xd.lag
Call:
lagsarlm(formula = X1 ~ X3, data = xd, listw = xd.nblsw, tol.solve =
1e-12)
Type: lag
Coefficients:
rho (Intercept) X3
0.012966193 -0.184136343 0.001995207
Log likelihood: 2.532483
>
Note that the results include estimates of the coefcients (intercept and slope) and the parameter
rho for ρ. This model is then used to predict the X1 rate by region. It was necessary to decrease the
tolerance to 10
−12
. By de ...