I am estimating a Weighted Spatial Simultaneous Autoregression Model (spdep::spautolm --> Link) in R and I would like to do some residual analysis.

Unfortunately functions such as hatvalues,cooks.distance or plot.lm do not work for spautolm objects. Yet, I would like to calculate leverages and cook distances for my model (see also my post on stackoverflow).

My Model looks like this:

$Y = X^T\beta + \lambda W(Y − X^T\beta)+ \epsilon$ with $\epsilon\sim^{iid} N(0,\sigma^2)$


$Var[Y]=\Sigma_{SAR} = \sigma²(I-\lambda W)^{-1}V(I-\lambda W)^{-1}$ with $V=diag[1/n_i]$

$\rho$ is my spatial autoregression parameter and $W$ the matrix that represents spatial dependence.

Obviously, a simple calculation of the hatmatrix $H$ via $H=X(X^TX)^{-1}X^T$ is not valid here due to the weighting and the modeled spatial autocorrelation.

Any ideas how to calculate the leverages and cook's distances by hand in R for this model?


Your Answer

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

Browse other questions tagged or ask your own question.