Using census data to investigate the multilevel structure of local populations
Using census data to investigate the multilevel structure of local populations
A methodology is described for the analysis of census data for a local population, so that the structure of this population may be investigated. In particular, it is possible to assess the (co)variation in socio economic variables at different levels of the population structure. The 1991 UK census data that are available for a local census population are considered, and it is explained how aggregate Enumeration District (ED) and ward level data from the Small Area Statistics (SAS) may be combined with individual level data from the 2% Sample of Anonymised records (SAR). These data allow a three level nested population structure to be considered, with a ward level, an ED level and an individual level.
Multilevel statistical models are specified, which reflect the 3 level nested population structure. Two types of model are considered: an unconditional (variance components) model (Model A) and a model conditional on a set of 'grouping variables' - variables which are thought to characterise population structure (Model B). Model A allows the extent of variation at different levels of the population to be estimated and Model B allows this variation to be explained, to some extent. It is explained how estimates of parameters may be obtained for Models A and B, even though the available data do not allow a 'standard' multilevel analysis to be carried out. Two approaches are described: a moments method and an Iterative Generalised Least Squares (IGLS) method. The parameter estimates (and functions of them) allow the extent of homogeneity within the EDs (and wards) to be estimated, and hence the variation at each level to be assessed. The methodology is successfully applied to 1991 UK census data, and the two approaches are compared via a simulation study. An important special case, where a level is ignored in the analysis is later considered in detail.
University of Southampton
1999
Tranmer, Mark David
(1999)
Using census data to investigate the multilevel structure of local populations.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
A methodology is described for the analysis of census data for a local population, so that the structure of this population may be investigated. In particular, it is possible to assess the (co)variation in socio economic variables at different levels of the population structure. The 1991 UK census data that are available for a local census population are considered, and it is explained how aggregate Enumeration District (ED) and ward level data from the Small Area Statistics (SAS) may be combined with individual level data from the 2% Sample of Anonymised records (SAR). These data allow a three level nested population structure to be considered, with a ward level, an ED level and an individual level.
Multilevel statistical models are specified, which reflect the 3 level nested population structure. Two types of model are considered: an unconditional (variance components) model (Model A) and a model conditional on a set of 'grouping variables' - variables which are thought to characterise population structure (Model B). Model A allows the extent of variation at different levels of the population to be estimated and Model B allows this variation to be explained, to some extent. It is explained how estimates of parameters may be obtained for Models A and B, even though the available data do not allow a 'standard' multilevel analysis to be carried out. Two approaches are described: a moments method and an Iterative Generalised Least Squares (IGLS) method. The parameter estimates (and functions of them) allow the extent of homogeneity within the EDs (and wards) to be estimated, and hence the variation at each level to be assessed. The methodology is successfully applied to 1991 UK census data, and the two approaches are compared via a simulation study. An important special case, where a level is ignored in the analysis is later considered in detail.
This record has no associated files available for download.
More information
Published date: 1999
Identifiers
Local EPrints ID: 463905
URI: http://eprints.soton.ac.uk/id/eprint/463905
PURE UUID: 7ee95ee8-c6bf-4d9e-94d1-2ad5858ba9fc
Catalogue record
Date deposited: 04 Jul 2022 20:58
Last modified: 04 Jul 2022 20:58
Export record
Contributors
Author:
Mark David Tranmer
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics