| dc.description.abstract | As an employee you can have a variety of insurances through your employer, one of them being life insurance covering death due to non-occupational illnesses. With such covers it is essential for the insurance company to know which factors impact risk and through this be able to predict the future risks for new policies. As companies enter and leave the portfolio from year to year, it induces shifts in the insured population and with that shifts in the observed death rates. This complicates the modelling of the death rates. In this thesis, we consider Generalized Linear Models (GLMs) and Generalized Additive Models (GAMs) for prediction and smoothing of nonlinear death rate patterns. We will consider different costumer properties for modelling and discuss differences and similarities in smoothing and predictions done by GLMs and GAMs. We, of course, find that death rates due to non-occupational illnesses increase with age. We also find that the death rates decrease over time. We detect significant differences in death rates of people working in companies with different NACE-codes, also known as activity codes. This is a mandatory statistical classification of the economic activities of a company, put down and regulated by the European Union. Here one of the more surprising discoveries is a higher death rate for women engaging in financial and insurance activities, compared to women in other NACE-codes tested, which usually require less education. | eng |