I have some data in a panel format and want to run some spatial panel regression. In a cross-sectional framework, I would run a simple regression and test the spatial autocorrelation from the residuals (with Moran's I for instance). If the residuals are spatially correlated, that would suggest me to try a spatial regression.

How can I do it on with a panel data?

I've found just one R package that performs this, but it seems weird to me as it's not possible to choose the spatial weights matrix. From what I get we can just use a contiguity one.

Is this the right way to test for spatial correlation on the residuals of a panel estimation? Is there any other package (R or another language) to do this?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.