333 – Quality of Alternative Sources for Social, Economic, and Health Data
Progressive Data Modelling
Zahoor Ahmad
University of Southampton
Li-Chun Zhang
University of Southampton
Currently a substantial initiative is taken at the National Statistical Offices to make full use of administrative data in statistical production. For example, several studies have previously been carried out at the United Kingdom Office of National Statistics (ONS), such as forecasting value-added-tax (VAT) turnover at the unit-level, adjusting VAT register totals towards the existing Monthly Business Survey (MBS) -based turnover estimates etc. The VAT data are said to be progressive in the sense that VAT reports (or observations) of a particular time period t of interest may arrive at various points long after t. For timely prediction of the VAT turnover total before all the data have arrived, a critical issue is when the timeliness of VAT reporting is related to VAT turnover i.e. informative reporting. In this work we develop new approaches for handling informative reporting, drawing on the relevant techniques for informative sampling and informative nonresponse. We study approaches to modelling the potential informative report, methods for estimation, including maximum likelihood and estimation equations, and illustrate the methodology.