Our department is known for its research in small area estimation, a suite of methods for making estimates from survey and related data sources, when only a small amount of survey data is available. Statistical models borrow strength from related variables, improving the quality of estimates for small domains. Our researchers led the Innovations in Small Area Estimation Methodologies project and were involved in Borrowing Strength, both funded by the Economic and Social Research Council.
Methodological developments and applications in this research group cover a wide range of small area estimation approaches, including area- and unit-level models, structure preserving estimating equations (SPREE), and state space models. The dependence of the approaches on models makes the use of robust estimation important for accuracy, and research topics include M-quantile and quantile regression and robust estimation for random effects models. These have been applied, for example, to estimate turnover in business surveys and extreme poverty using ‘big’ data sources (satellite and mobile phone data). Robust estimation also has wider applications, and recent research includes analysis of longitudinal data from the Millennium Cohort Study.