Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.
Author ORCID Identifier
AccessType
Open Access Dissertation
Document Type
dissertation
Degree Name
Doctor of Philosophy (PhD)
Degree Program
Public Health
Year Degree Awarded
2020
Month Degree Awarded
February
First Advisor
Leontine Alkema
Subject Categories
Biostatistics | Multivariate Analysis | Statistical Models | Vital and Health Statistics
Abstract
Population level mortality data is often subject to substantial reporting errors due to misclassification of cause of death, misclassification of death status, or age reporting errors. Accuracy of error-prone data sources can be assessed by comparing such data to gold standard data for the same population-period. We present Bayesian methods for assessing the extent of reporting errors across different population-periods and generalizing those to settings where gold-standard data are lacking. Firstly, we investigate misclassification errors of maternal cause of death reporting in civil registration vital statistics data. We use a Bayesian hierarchical bivariate random-walk model to estimate country-year specific sensitivity and specificity in countries with at least one period where vital registration data overlaps with gold standard data. For countries without gold standard data, we developed a sequential approach, in which fixed global estimates of sensitivity and specificity are used. Additionally, we propose a new approach to incorporate temporal structure of misclassification parameters. Secondly, we investigate misreporting of adult mortality in sibling survival history data. Sibling survival histories data suffers from reporting errors due to respondent misreporting of birth year and age at death of their maternal siblings. We perform an exploratory analysis of data collected in Malawi and propose a candidate parametrization for reporting errors in cohort survival probabilities by 5-year age groups. We introduce parameters to capture age-group specific age-at-death errors and birth year reporting errors and define the data generating processes that relate sibling survival data to true survival probabilities while accounting for reporting errors. This framework allows for the estimation of age-group specific survival probabilities in settings where only error-prone sibling survival history data is available.
DOI
https://doi.org/10.7275/15997788
Recommended Citation
Peterson, Emily, "BAYESIAN METHODS FOR THE ASSESSMENT OF REPORTING ERRORS FOR DATA-SPARSE POPULATION-PERIODS WITH APPLICATIONS TO ESTIMATING MORTALITY" (2020). Doctoral Dissertations. 1861.
https://doi.org/10.7275/15997788
https://scholarworks.umass.edu/dissertations_2/1861
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Included in
Biostatistics Commons, Multivariate Analysis Commons, Statistical Models Commons, Vital and Health Statistics Commons