[Openspace] Potential Two-Step Approach

[kwhite at ssc.wisc.edu]

I'm new to this list and have a question pertaining to a potential
two-step approach to address spatial autocorrelation using GeoDa.

The situation: I have time series data of county units (ten time points). 
Diagnostics of the OLS suggest that both a spatial lag and a spatial error
model are appropriate; one is usually "better" but the scores are
remarkably similar.  Further, the diagnostics for the respective spatial
regression technique suggest that the other form persists.

The potential solution: Would it be appropriate to run a spatial lag model
with all of the covariates and save the residuals, and then run a spatial
error model with all of the covariates and the residuals yielded from the
spatial lag model?

Thanks in advance for your insights.

Katherine