Eigenvector Filtering (EF) is a popular technique for the analysis of spatial or space-time lattice data. In spatial data analyses, EF is based on synthetic predictors which represent distinct map patterns; these control variables capture stochastic spatial dependencies in the residuals, allowing model building to proceed as if observations were independent. In this talk I will discuss how EF compares to well-known spatial econometric models and Geographically Weighted Regressions. A new algorithm for EF filtering will be presented; the algorithm combines variance-inflation-factor (VIF) filtering, correlation-based screening and a fast adaptive lasso estimator with EF. The new scheme possesses a number of advantages: it can be implemented to estimate challenging specifications (e.g. space-time models with spatially varying coefficients) and frequently it is fast enough to allow for bootstrap confidence intervals. In a series on Monte Carlo experiments, the proposed method is compared against existing Hierarchical Bayesian specifications and conventional forward selection schemes for EF. Finally, an application related to the diffusion of innovation in US agriculture will be briefly discussed.