[Zurück]

M. Weinauer:
"Analysis and forecasting of Austrian mortality data with methods of compositional data analysis";
Betreuer/in(nen): M. Templ; Instituts für Stochastik und Wirtschaftsmathematik, 2017.

Forecasting cause-specific mortality helps to estimate future health expenditure and to plan spending on research, capital investment, preventive measures or palliative care for sub-groups. The objective of this thesis was to analyse past trends (1983-2015) of the mortality for 73 causes of death as well as to forecast mortality for these 73 causes in Austria in the future (2030). To assure for coincidence of aggregated forecasts of single death causes with forecasts of total death causes, a compositional extension of the well-known Lee-Carter model -- the Oeppen model -- was applied for forecasting. For analysis of the past the compositional counterparts of methods of principal component analysis (PCA) were applied: Compositional PCA and compositional sparse PCA were used to reveal time trends, simplicial PCA to discern age patterns. Calculations were executed in R and Matlab. As in the past, also in 2030 diseases of the circulatory system constitute the most dominant cause of death. Death counts from the ICD-10 classification "other diseases" as for example diabetes mellitus and Alzheimer´s disease will rise markedly until 2030. As these diseases are directly linked to lifestyle, the prediction reflects the Austrian change of lifestyle towards little physical activity and unhealthy diet. Other causes as asthma or intentional self-harm are projected to decrease.

Mortality Modelling, Compositional Data Analysis

http://katalog.ub.tuwien.ac.at/AC14507176