Author: Partha Lahiri (University of Maryland)
Due to advances in computing, government agencies can process administrative records and link them with sample survey and census records for statistical purposes in a fraction of time and costs required for field data collection. The accessibility of different administrative data from different sources has brought new opportunities for statisticians to develop innovative small area estimation (SAE) methods that can cut down costs and improve the quality of estimates. For example, a survey and an administrative record source could cover some common portions of a population of interest. Inevitably, record linkage can produce errors either by missing matches or by erroneously linking two different people together. Assessing uncertainty in small area estimation when at least a portion of the data used for estimation arises from a record linkage procedure is a novel problem. In this talk, I will present a general theory for small area estimation for probabilistically linked data.