Pembangunan model regresi poisson sifar-melambung berintegrasi dalam biostatistik

• Main Author: Zafakali, Nur Syabiha
• Format: Thesis
• Language: English
• Published: 2018
• Subjects: RA0421 Public health. Hygiene. Preventive Medicine
• Online Access: http://eprints.usm.my/47234/1/Dr.%20Nur%20Syabiha%20Zafakali-24%20pages.pdf; http://eprints.usm.my/47234/
• Institution: Universiti Sains Malaysia

The Poisson regression method is an approach often chosen by researchers in making analysis of data in the form of numbers. However, the excess zero often applies to such data. This is due to the existence of overdispersion of the data collected. This phenomenon led to the Poisson regression model is no longer suitable to be applied. Alternatively, the method of Zero-Inflated Poisson regression was selected to model the data that have excess zero phenomenon (overdispersion). This research also emphasizes the development methodology of data analysis. The data gathered involved two major cases studies of data from patients with Thalassemia among children and the patients with dental caries problems. The first phase in this research is to refer to the algorithm development procedure to model the Zero-Inflated Poisson Regression method through the bootstrap method and combined with the fuzzy regression method. The combination of these methods is referred to as the Integrated Model. The second phase is the comparison of the findings between the Integrated Model and the existing method. An overview of the overall model was also performed to obtain information related to the efficiency of the model. The algorithm was developed based on the concept of improvement and every detail of the methods used will be explained carefully on the methodology. The main outcome of this research is to refer to the development of a research methodology that also helps researchers to analyse data more effectively and to give more accurate results. The use of the Integrated Model of the two case studies has resulted in a more efficient average value compared to the existing method. It shows that this method has demonstrated a better model for each set of data that can be studied and applied successfully.