![IconGems-Print](images/IconGems-Print.png)
252 – SPEED: Nonparametrics and Imaging
Estimation of Locally Stationary Spatial Processes with Application to the American Community Survey
Daniel Weinberg
U.S. Census Bureau
Tucker McElroy
U.S. Census Bureau
Soumendra Lahiri
North Carolina State University
The American Community Survey (ACS) multiyear estimates provide detailed economic and demographic information at a census tract level. The assumption of spatial stationarity for many variables is dubious, which motivates our formulation of local stationarity that can take into account changes in the covariance structure across census tracts. In addition, we adopt a nonparametric modeling approach that remains agnostic about specific distributional features. We present fairly general constructions in both the frequency and spatial domains, deriving an estimator for the local covariance. The properties of the local covariance estimator are explored through simulation. For our application, we utilize our estimator on the ACS data of median household income in the state of Iowa.