Transfer effective-sample-size correction to geographically weighted regression

Determine whether an effective-sample-size argument can be transferred to geographically weighted regression, where location-specific coefficient estimates create a multiple-testing problem and neighboring local regressions share observations.

Background

The paper develops an effective-sample-size adjustment for Bonferroni and Benjamini–Hochberg corrections applied to Local Indicators of Spatial Association (LISA), accounting for the fact that spatial dependence reduces the amount of independent information relative to the nominal number of spatial units. The authors note that geographically weighted regression produces one set of coefficients per location, creating a related multiple-testing problem.

Because neighboring geographically weighted regressions also share observations by construction, the dependence structure may differ from that of LISA tests. The paper identifies the transferability of the effective-sample-size principle to this setting as an unresolved question, beyond the analyses conducted in the paper.