Analysis of high-dimensional data
- Objective: New medical measuring devices generate high-dimensional functional data, but the knowledge on how to analyse such data is limited. Functional data analysis (FDA) is useful for data that are not measured as functions, but has an underlying functional nature. Analysis of longitudinal data is important, but difficult. Change can take many forms, it may be abrupt or occur more smoothly over time. It may be influenced (in an unknown manner) by a number of factors which themselves may vary with time, as might the manner in which they influence change.
Atrial fibrillation in a cohort of 432.000 Norwegian men and women
- Objective: to estimate the direct and indirect effects of leisure time physical exercise on risk of atrial fibrillation in a general population that includes both genders based on a causal model for atrial fibrillation in which potential confounding factors can be accounted for.
- Objective: Boosting was originally put forth as a classification algorithm, but has subsequently been extended to a number of important problems statistics. We have applied Boosting for ME, and will extend this to GLMM and longitudinal analysis.
Changes in causes of death during the twentieth century
- Objective: to describe changes in causes of death in Norway 1951-2007.
Exact statistical analysis of 2x2 contingency tables with small cell counts
- Objective: Statistical analysis of 2x2 contingency tables is usually done by Pearson’s chi-squared test and Fisher’s exact test, and confidence intervals are mostly calculated by using asymptotic theory. We study the efficiency of the methods, with specific focus on tables with small cell counts.
Follow-up of the Tromsø Heart Studies
- Objective: to analyze the changes in gender and age differences regarding trends in coronary heart disease over three decades, 1974-2004.
Life style and biological factors on mortality (42 year follow-up)
- Objective: to assess the effect of physical activity in leisure and mortality in the elderly- follow of the Bergen Study.
Measurement error modelling
- Objective: Our interest in ME is on correction methods, since the regression between the outcome variable and the observed and error contaminated explanatory variables is a diluted version of the relationship between the outcome and the corresponding error free explanatory variables. Ongoing work includes studies of the Cochran-Armitage test for trend.
Protective effect of alcohol on cardiovascular diseases
- Objective: to assess whether the cardio-protective effect of alcohol is mediated via high density lipoprotein cholesterol - A prospective cohort study.
Quantile regression analysis
- Objective: The distance function between cumulative distributions, given as the difference between their inverses, has shown to be of great interest for studying variability. We extend the estimation of a distance function to take into account a set of explanatory variable that might influence our variable of main interest. Estimation is done by Boosting.
The association between life expectancy and education
- Objective: to assess trends in life expectancy by education in Norway 1961-2008, a register-based population study.
The interaction between susceptibility genes, environments and life style as disease causes
The Intergene program, a study of the INTERplay between GENEtical susceptibility, environmental factors including life-style and psychosocial background for the risk of cardiovascular diseases in South West Sweden.