Speaker
Bo Henry Lindqvist, Professor, Department of Mathematical Sciences, Norwegian University of Science and Technology
Abstract
The background for the present talk is a student project analysing data from HUNT (HelseUndersøkelsen i Nord-Trøndelag). This is an extensive health survey which is composed of three studies, HUNT 1 (1984-1987), HUNT 2 (1995-1997) and HUNT 3 (2006-2008). In the present project 74943 participants in HUNT 1 (which included more than 80% of the population in Nord-Trøndelag) have been followed up until 2004 with respect to death and cause of death. Altogether 22284 of the participants had died at the end of 2004.
The main objective of the study was to investigate the effect of different risk variables on a set of death causes, such as stroke and heart failure. The talk will discuss various aspects of regression modelling of both cause specific hazards and cumulative incidence probabilities, where the traditional approaches are using, respectively, ordinary Cox regression and the "Fine-Gray" setup. Modified approaches which take into account the large size of the risk sets and the relatively few cases for each death cause, will in particular be considered in the talk.