- Title
- Application of some missing data techniques in estimating missing data in high blood pressure covariates
- Creator
- Odeyemi, A. S
- Subject
- Missing observations (Statistics) Hypertension
- Date Issued
- 2019
- Date
- 2019
- Type
- Thesis
- Type
- Doctoral
- Type
- PhD
- Identifier
- http://hdl.handle.net/10353/15189
- Identifier
- vital:40195
- Description
- Cases recorded with high blood pressure are a major concern in both public and private hospitals. Adequate provision of health information of patients relating to high blood pressure in Eastern Cape Hospitals hinges so much on the outcome of statistical analysis results. The usual statistical methodologies become inadequate in handling statistical analysis of data collected due to incomplete patients’ information stored in the hospital database. From time to time, new methods are developed to address the problem of missing data. High blood pressure is linked to a lot of diseases such hypertension, cardiovascular disease, kidney disease and stroke. In this study, we developed a new method for addressing the problem of missing data in assessing model used for estimating missing values in terms of minimum errors(using RMSE, MAE, and SE) and goodness-of-fits(using 2 R and adjusted 2 R ) of this model and P-value. . The study compared six different methods: Original data (OD), Listwise deletion (LD), Mean imputations (MEI), Mean above (MA), and Mean above below(MAB) and two steps nearest neighbour (2-NN).The comparison was performed using original data set, and missing values at 5%, 10%, 20%, 30% were simulated on Framingham risk scores under MCAR and MAR simulation on BMI values given some assumptions. Five performance indicators were used to describe the model minimum errors and goodness of fit for all the methods. The results showed that the 2-NN is the best replacement method at lower levels (5% and10%) of missing values while MA and MEI performed best at higher levels(15% and 20%) of missing values. All comparison was based on estimates closest to those of the original data where no value was missing. MAR results showed that 2-NN performed better than LD,MA,MAB, and MEI at 5%,10%, and 20% levels of missing data in terms of absolute difference in p-value to original data.
- Format
- 149 leaves
- Format
- Publisher
- University of Fort Hare
- Publisher
- Faculty of Science and Agriculture
- Language
- English
- Rights
- University of Fort Hare
- Hits: 593
- Visitors: 528
- Downloads: 51
Thumbnail | File | Description | Size | Format | |||
---|---|---|---|---|---|---|---|
View Details Download | SOURCE1 | Odeyemi AS PhD Thesis 2019.pdf | 3 MB | Adobe Acrobat PDF | View Details Download |