Absolute risk estimation for time to event data

Abstract

The objective of the thesis is to use time to event models in order to estimate the absolute risk for a certain event. In particular, we will use the data from the Estonian Biobank cohort together with different approaches to estimate the Risk of Type 2 Diabetes (T2D). We will use the methodology that accounts for right-censoring in the data. Specifically, we will use three approaches for duration models: - Non-Parametric methods: the Kaplan-Meier estimator; - Semiparametric model: Cox Proportional Hazard models; and - Parametric models: models assuming Weibull and Gompertz distribution. The analysis will be done in R software exclusively. After we have identified the optimal models, we will predict the risks, giving us an approximate estimate which will be potentially useful to personalize risk predictions for the Estonian population and insurance.