In 2007, the Department of Housing and Urban Development initiated a point-in-time count of the homeless across the United States. The counts are administered by the Continuum of Care Program, which provides spatial and temporal data for the homeless population over the last decade. Unfortunately, this administrative spatial unit does not align with the more common areal units defined by the United States Census Bureau, which limits usability of these data. To unify these two areal units, spatial disaggregation, matching, and imputation allow for aligning Continuum of Care data with county data. The resulting county-level homeless counts for the years 2005 to 2017 are provided as an R package. The county-level data display more spatial precision and more temporal variation than the Continuum of Care-level data. Nonparametric regression analyses reveal that the spatiotemporal variation in the data can be well approximated by additive spatial and temporal effects at both the county and Continuum of Care level.
Bibliographical notePublisher Copyright:
© 2019, © 2019 Taylor & Francis Group, LLC.
- small area estimation
- spatiotemporal modeling