A0525
Title: Inference for eigenvector spatial filtering
Authors: Rowan Cherodian - University of Sheffield (United Kingdom) [presenting]
Guy Tchuente - Purdue University (United States)
Sylvain Barde - University of Kent (United Kingdom)
Abstract: The post-selection inference problem is investigated in the context of eigenvector spatial filtering (ESF). ESF uses a subset of eigenvectors from a spatial weights matrix to efficiently account for any omitted cross-sectional correlation terms in a classical linear regression framework; thus, it does not require the researcher to explicitly specify the spatial part of the underlying structural model. The conditions necessary for consistent and asymptotically normal parameter estimation are shown, assuming the support (relevant) set of eigenvectors is known. Several methods have been proposed to select the relevant sub-set of eigenvectors. A recent study was the first to propose a procedure that accounts for the post-selection inference problem. The simulation results also show that the partial regression type procedure proposed in the referred study can be extended to other ESF selection procedures to construct valid confidence intervals. This indicates that their partial regression procedure is valid across ESF selection procedures.
