Statistical and computational issues in fitting a new model for indigenous life expectancy
Date
2014
Authors
Fu, Lulu
Journal Title
Journal ISSN
Volume Title
Publisher
Canberra, ACT : The Australian National University
Abstract
The Australian population experiences the lowest mortality in the world. However, mortality rates for Indigenous Australians are significantly higher than for the non-Indigenous population across all age groups. Consequently, there is a huge gap between Indigenous and non-Indigenous life expectancy. The gap between Indigenous and non-Indigenous life expectancy is of great concern to the Australian government. In 2008, the Council of Australian Governments set a goal to close the gap by 2031. Although life expectancy plays a crucial role in assessing Indigenous disadvantage, estimation of Indigenous life expectancy is particularly difficult. Both Indigenous death registrations and population estimates are subject to incompleteness and accuracy issues. A main factor causing these issues is difficulties in correctly identifying Indigenous status. This is because a person's Indigenous status could be either misidentified or not identified at all. Many life expectancy estimation methods have been developed to address the accuracy and completeness issues of death registrations. The most well-known methods are the Brass method, the Hill method, the Bhat method and the record linkage method. However, each of the existing methods has some limitations. The Brass method, the Hill method and the Bhat method ignore people with unidentified Indigenous status. In addition, the Brass method and the Hill method do not take into account migration. For the Bhat method, the Australian Bureau of Statistics (ABS) discovered that it can produce results that are inconsistent with observed data. The record linkage method is able to incorporate people whose Indigenous status is unknown but it is very complicated to apply. The method requires accurate and confidential data such as name, age, sex and place of usual residence. Such detailed data may not be readily available or the data may lack accuracy. In addition, all these methods rely on reported Indigenous status (either from the individual themselves or from a relative), which could be unreliable or inconsistent. In this thesis, we propose a new mathematical model and explore the possibility to overcome the problems associated with the estimation of Indigenous population life expectancy. The model does not require detailed personal information. It is very flexible and allows the possibility that a person's Indigenous status is wrongly reported or unidentified. However, it is quite computationally demanding to implement. For the purpose of this thesis, we focus on the statistical and computational issues in fitting the model and investigate various strategies to increase the feasibility of actually using the model to produce reliable estimates.
Description
Keywords
Citation
Collections
Source
Type
Thesis (PhD)
Book Title
Entity type
Access Statement
License Rights
Restricted until
Downloads
File
Description