Social Statistics Seminar - 4pm, 1 December: Dual and multiple system estimation with partially and imperfectly observed covariates
|Dates:||1 December 2020|
|Times:||16:00 - 17:00|
|What is it:||Webinar|
|Organiser:||Cathie Marsh Institute for Social Research|
Social Statistics Seminars 2020/21 – December Session
Dual and multiple system estimation with partially and imperfectly observed covariates
Paul A. Smith (University of Southampton)
Join us on Zoom!
Topic: Social Statistics Seminars 2020/21 – December
Time: Dec 1, 2020 04:00 PM London
Meeting ID: 924 2925 3808
Dual (and multiple) system estimation, also known as capture-recapture, has become widely used in official statistics for estimating the size of partly observed populations, with applications from population censuses to linked administrative data. There are situations in which sources can be linked, allowing estimation through a loglinear modelling procedure, but where some variables appear in only one source. This additional missingness can be handled using the EM algorithm, and we illustrate its application. Sometimes the data to be linked contain what should be the same variable, but measured in different ways, and we can use the same approach to analyse and adjust the data to the preferred measure. In more complex situations with many inputs it is possible to produce a consistent estimate from the disparate definitions, and we illustrate this with recent work on estimating the size of the Maori population in New Zealand.
About the speaker
Paul Smith is Professor of Official Statistics at the University of Southampton, with a particular interest in sampling, estimation and other statistical methods applied to government statistics. He worked for many years in the Office for National Statistics (ONS) before moving to academia, which has provided a solid fund of challenging research ideas, and he continues to work with the ONS on a range of different topics. He is just finishing a term as Discussion Papers Editor for the Journals of the Royal Statistical Society.
Travel and Contact Information