| Summary: | This thesis uses data from the Tromsø Study and weather data from the Norwegian Meteorological Institute to study the mortality hazard rate and incidence rate of myocardial infarction (MI) in Tromsø using a latent Gaussian modelling framework. Inference is performed using integrated nested Laplace approximations (INLA). This thesis presents the datasets and describes the modelling and computational framework, before analysis is performed. To study the mortality hazard rate after MI, a Cox proportional hazards model has been implemented. A model without a seasonal effect with sex as a stratum variable was deemed the best fit. The results show an increased risk in the month after a MI. After the first month, the risk drops, before it increases with age. The mortality hazard rate is slightly higher for men than for women. To study the change in the rate of MI during the time of the study, a Bayesian age-period-cohort model has been implemented. The model only includes the men of the study. This model studies the rate of MI on three different time scales: age, period, and cohort. The effects from age, period, and cohort are not directly identifiable. However, second differences describing the curvature and relative risk ratios are identifiable, as is the overall rate. The results show the incidence rate of MI decreasing with period, and increasing with age.
|
|---|