Spatial Correlation Robust Inference

We propose a method for constructing confidence intervals that account many forms of spatial correlation. The interval has the familiar “estimator plus and minus standard error times critical value” form, but we new methods value. is constructed using population principal components from given “worst‐case” correlation model. value chosen to ensure coverage in benchmark parametric model correlations. shown control finite sample Gaussian settings restricted nonparametric class models large samples whenever weak, is, with average pairwise correlations vanish as size gets large. also provide results on efficiency method.