Date: Tuesday, 21 November 2023, at 12:15 pm
Venue: Room Dono Giannessi, DEM
Speaker and Title:
María Bugallo Porto (
Temporal M-quantile models for Small Area Estimation
Sample surveys are extensively used to collect data for calculation of reliable direct estimates of population parameters. Moreover, reliable estimates for sub-domains are also required to implement well-targeted policies. However, if sample sizes are small, direct estimators may be unreliable. This problem can be avoided by employing small area estimation (SAE) techniques. The nested error regression (NER) model is considered the basic unit-level model in SAE. However, its nesting structure is unable to model time-dependent data when the number of time periods is somewhat large. Nevertheless, in many cases it is reasonable that observations close in time are more similar than those further away. Borrowing strength from time is therefore a smart strategy. In addition, we are not aware of published studies that address robust prediction in small areas based on time-dependent data. Thus the aim is to extend the classical M-quantile (MQ) model to this field of research. At this regard, the proposed Time-weighted M-quantile (TWMQ) model successfully addresses this challenge and offers flexibility to the widely imposed assumption of unit-level independence. An application to income data in Spain is proposed.
Discussant: Gaia Bertarelli (Università Ca’ Foscari Venezia)