Fellows Research Meetings

Stephen Horn
Chair, Official Statistics Section, SSA

On surveys in Official Statistics: State spaces or space stations

Surveys have been the mainstay of official statistics. In their current model-assisted garb surveys as a vehicle for official statistics have been overtaken in several important respects. Challenge from alternative sources has seen the inference base for official series suddenly vulnerable. The new data is immediate, dynamic, not tied to sampling for identification, yet incomplete and of varying scope and quality: the international integration project has parted ways from the developments in survey theory on which confidence in official statistics had rested.

I outline a theoretical response to this challenge that can restore a base to survey inference even as agencies abandon design base to capture value from newly trafficable data stores. Confidence in the path proposed comes from its theoretical underpinning as a natural extension of model assistance; in early results from full scale testing of associated algorithms; and from its links to applications in inference over time, small areas and multi-source linkage. This talk is a progress report of my long term collaboration with Dr Raymond Czaplewski, statistician emeritus with the US Forestry Service.

Last reviewed: 15 December, 2017