Abstract:
|
Statistics at detailed levels of aggregation can inform data-driven policy making, among many other benefits. However, sample surveys are typically not designed to provide quality inference for all quantities of interest (e.g., fine geographic and/or demographic levels). Small area estimation involves exploiting relationships among domains and borrowing strength from multiple sources to improve inference. This typically involves the use of models whose success depends heavily on the quality and predictive ability of the sources used. Successful models often lead to dramatic reductions in uncertainty measures, and can support inference for areas with no sample. One rich source of information is that of other surveys, especially in the US, where multiple surveys exist that cover related topics. Other sources include administrative records, Censuses, big data such as traffic or cell phone data, or previous vintages of the same survey. We will discuss the practical and statistical challenges of combining information from various sources in small area estimation, encouraging participants to share their own experiences, research, and the practical challenges they have faced.
|