Modelling of Health Data using Time Series Models

Abstract

This research study is focused on decoding the time series data that is continuous into ordinal data that changes over time, using mixture distributions and transition probabilities using hidden Markov model (HMM). The information that has come out of HMM were used to address, the duration of stay of an epidemic and the likely time to move from one stage of an epidemic to another using cholera data over 15 years. Cholera is an infectious disease and hence the monthly cholera count has a dependency nature among them. The Bayesian approaches are also used in the analyses, as the prevalence of severe form of cholera was small. The monthly count data on cholera infections (transiting mixtures) was further analysed using Markov regression and Bayesian ordinal logistic regression analyses. The results of risk factors were comparable to each other. Thus this dissertation established a model to analyze the data beyond time series data and Poisson regression methods to Markov regression. From the mean passage time estimated it is clear that the average time for the occurrence of a higher number cholera counts from a less occurrence time period is approximately one year. And the expected time to go back to a low level occurrence alter observing the higher occurrence is nine months, Once the higher count occurred it will stay for around three months and but the lower level occurrence of disease is for 2 to 4 months only. The duration of stay in the low and moderate levels cholera counts are 5 months and 3 months respectively. In short, the seasonality and the influence of climate for this higher cholera incidence every year reflects in these findings. The monthly caesarean counts are always independent and also over-dispersed. It was assumed that they do not satisfy Markovian principles and therefore there is a need to handle them in understanding the over-dispersion. ARIMA models are always appropriate for the forecasting or prediction analysis. The primary caesarean proportion is increasing each year and it reaches 22% in 2010.