Analysis of hospital admissions due to respiratory diseases between 2010 to 2017 in Hong Kong

Abstract

Hospital admissions due to respiratory diseases (HARD) has been widely discussed in the past three decades, and has been linked to air pollutants, media information, and dynamic weather conditions, which can be observed daily and act as smooth trajectories. Classical research works mainly analyzed HARD through cross-sectional studies. It is also interesting and challenging to detect the effects of environmental conditions on HARD from the new perspective of functional data analysis. Motivated by the aforementioned problem, the thesis aims to two targets based on functional data analysis: one is to diagnose the risk of HARD through media information and weather conditions; the other is to explore how air pollutants and weather conditions impact on HARD through a new functional regression model. Part I focuses on improving diagnosis of high-or low-hospital admissions by combining media information with weather conditions, the multiple functional markers. There is rich literature in combining scalar markers to improve diagnostic accuracy, but they are inapplicable for functional markers. We propose a scalar feature to represent the original functional curve, so that existing scalar combination methods can be applied. Part II tries to explore a new functional additive regression model to characterize the complicated influence of weather conditions and air pollutants on HARD. I suggest some estimation procedures for the coefficients in the new functional model with hospital admissions as response. Such investigation from functional data analysis perspective can be also applied to other real worlds data problems that have intense daily records over many years.