- Title
- Bayesian spatio-temporal zero-inflated mixed models for overdispersion on chronic disease mapping
- Creator
- Osuji, Georgeleen O https://orcid.org/0000-0002-8408-3928
- Subject
- Medical mapping
- Subject
- Bayesian statistical decision theory
- Date
- 2021-12
- Type
- Doctoral theses
- Type
- text
- Identifier
- http://hdl.handle.net/10353/23644
- Identifier
- vital:58230
- Description
- Background: Life expectancy in most developing countries has remarkably increased and decreased in mortality, but under 5 years old mortality has increased due to HIV and Tuberculosis incidence. Many factors have been established to influence the mortality rate among HIV patients and understanding the factors contribution to the risk of under 5-year-old mortality is important for designing appropriate health interventions. Excess zeros usually occur in such HIV mortality count data. Mixed models consisting of count part and zero part are often used to describe the observed excess zero in the data. Poisson models are popular modeling inference, but Negative-Binomial models are more flexible in analyzing count data and dealing with overdispersion. Method: This research proposed to develop two-part hurdle models in analyzing areal zero count data. A spatial Bayesian lognormal-logit hurdle model (BLLHM) with random effects characterizes and cross-spatial dependencies were introduced. The parameter inferences and predictions were evaluated using the Markov Chain Monte Carlo algorithm. The model proposed was applied to HIV-positive under 5-year-old mortality collected from the Eastern Cape Department of Health. Results: Bayesian lognormal-logit hurdle model is selected as the best model fit. It is observed that the total number of HIV patients not on ART-HIVnotTB (0.000612, p <0.000) was positively and statistically significantly associated with the HIV-positive mortality of under 5 years patients. Both CD4 counts were done on newly diagnosed HIV rate (CD4count) and HIV-positive new patients screened for TB rate (HIVTBrate) were negatively and statistically significantly associated with the HIV-positive mortality of under 5 years patients (-0.6294, p = 0.000 and -0.00056, p = 0.0052). However, the covariate HIV positive Tuberculosis Preventive therapy (TPT) uptake rate (HIVandTB) was not statistically significantly associated with the HIV-positive mortality of under 5 years patients (-0.00155, p = 0.5392). Conclusion: The model is flexible to deal with zero-inflated and over-dispersed count data. There is a need to consider the risk of cause-specific under-5-year-old mortality in terms of spatial effects.
- Description
- Thesis (PhD) -- Faculty of Science and Agriculture, 2021
- Format
- computer
- Format
- online resource
- Format
- application/pdf
- Format
- 1 online resource (139 leaves)
- Format
- Publisher
- University of Fort Hare
- Publisher
- Faculty of Science and Agriculture
- Language
- English
- Rights
- University of Fort Hare
- Rights
- All Rights Reserved
- Rights
- Open Access
- Hits: 491
- Visitors: 510
- Downloads: 45
Collections
UFH Department of Statistics
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | SOURCE1 | Osuji-Full Thesis-PhD-BioStatistics.pdf | 2 MB | Adobe Acrobat PDF | View Details Download |