CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Small area estimation in the presence of data masking Authors:  Nikos Tzavidis - University of Southampton (United Kingdom) [presenting]
Abstract: The production of official statistics is explored, in particular, estimation for small geographic areas in the presence of data masking. The main focus will be on estimation when the response variable is grouped due to concerns about data confidentiality or survey response burden. Reporting data in groups (bands) is a mechanism that is employed, for example, in surveys collecting information on income. Methodology that enables fitting a random-effects model when the dependent variable is grouped will be outlined. Model parameters are then used for small area prediction of finite population parameters. Model fitting is based on the use of a stochastic EM (SEM) algorithm. Since the SEM algorithm relies on Gaussian assumptions, adaptive transformations are developed for handling departures from normality. The estimation of the mean squared error of the small area parameters is facilitated by a parametric bootstrap that captures the additional uncertainty due to the censoring mechanism and the use of transformations. The presentation will also briefly discuss the production of official statistics (a) when geographical coordinates are aggregated and (b) when data is geographically displaced.