NTNU, Trondheim, Norway
For proposals of bachelor/project/master-thesis topics please see here
Current PhD students
- Umut Altay (Co-supervisor) (02/2020 - )
Completed theses
PhD theses
- Ingeborg Gullikstad Hem (joint supervision with Geir-Arne Fuglstad) (08/2017 - 07/2021 ), Robustifying Bayesian Hierarchical Models Using Intuitive Prior Elicitation.
- Jingyi Guo (08/2013-08/2016), Bayesian Meta-analysis. Defended on: 06.12.2016.
- Alexander Knight (08/2015 - 2020) "Multilevel models and the development of new default priors based on penalised complexity priors". Terminated without graduation.
Master theses
- Vojin Kurtovic Dudic:
Analysis of German cancer data for lung-, prostate- and breast cancer (08/2020-10/2021)
- Nan Amalie Videng:
Joint spatio-temporal modelling of brain cancer incidence and mortality in Norway (01/2020-06/2020)
- Kristoffer Kofoed Rødvei:
Comparing the ACER and POT MCMC Extreme Value Statistics Methods Through Analysis of Commodities Data (08/2015-04/2016, co-supervisor)
- Maxime Conjard:
Towards joint disease mapping using penalised complexity priors (08/2015-07/2016)
- John Darkwah:
Bayesian inference for disease mapping: Comparing INLA and MCMC (08/2015-07/2016)
- James Korley Attuquaye:
The impact of varying time scales on the quality of cancer projections based on the Bayesian age-period-cohort model (08/2014-07/2015)
Bachelor theses
- Guro Aglen:
Hidden Markov models (01/2016-08/2016)
Full lectures:
- 01/2022-07/2022: MA8702 Advanced computer-intensive statistical methods
- 08/2021-12/2021: ST1201/ST6201 Statistical Methods
- 01/2021-07/2021: MA8702 Advanced computer-intensive statistical methods
- 08/2020-12/2020: ST1201/ST6201 Statistical Methods
- 08/2018-11/2018: ST1201/ST6201 Statistical Methods (last part was not given due to family health issues)
- 01/2016-07/2016: TMA4300 Computer intensive statistical methods
- 08/2015-12/2015: TMA4265 Stochastic processes
- 01/2015-07/2015: TMA4300 Computer intensive statistical methods
- 08/2014-12/2014: TMA4265 Stochastic processes
- 01/2014-07/2014: TMA4300 Computer intensive statistical methods
- 08/2013-12/2013: TMA4265 Stochastic processes
University of Zurich, Switzerland
Full lecture:
- 02/2012-06/2012: Bayesian inference (for M.Sc. in Biostatistics)
Exercise classes:
- 09/2010-12/2010: Bayesian inference (for M.Sc. in Statistics)
- 09/2009–12/2009: Biostatistics (for M.Sc. in Medical Biology)
Other:
- 01/2009–08/2009: Development of Java-applets to visualise statistical concepts
Integrated nested Laplace approxmiations (INLA) courses
- 2 day course (jointly with Haakon Bakka) University of Zurich, Switzerland, May 12-13, 2016.
Course description
Course material: See here
References for computing the continuous ranked probabiity score (CRPS): The definition is provided in the paper by Gneiting and Raftery, equation 20:
Gneiting and Raftery (2007), JASA, Volume 12, Number 102
If the predictive distribution can be assumed to be normal (which often applies approximately) a simplified formula, which is given directly below equation 20, can be used.
Here, only the mean and standard deviation of the predictive distribution are needed which are available from INLA.
Otherwise, the CRPS can be computed using a Monte Carlo approach based on samples (see inla.posterior sample example) from the predictive distribution see Equation 37
Sigrist et al. (2014), arXiv:1204.6118
- 3 hour course, Universiy of Toronto, Canada, March 27, 2016
- 2 day course, Universidad Publica de Navarra, Pamplona, Spain, December 1-2, 2015
- 3 hour pre-conference course, IBS channel network conference 2015, Nijmegen, Netherlands, April 20, 2015 Course material: see here
- 1.5 day course, University of Oslo, Norway, November 5-6, 201Course material: see here
Course material: See here
References for computing the continuous ranked probabiity score (CRPS): The definition is provided in the paper by Gneiting and Raftery, equation 20:
Gneiting and Raftery (2007), JASA, Volume 12, Number 102
If the predictive distribution can be assumed to be normal (which often applies approximately) a simplified formula, which is given directly below equation 20, can be used.
Here, only the mean and standard deviation of the predictive distribution are needed which are available from INLA.
Otherwise, the CRPS can be computed using a Monte Carlo approach based on samples (see inla.posterior sample example) from the predictive distribution see Equation 37
Sigrist et al. (2014), arXiv:1204.6118
Course material: see here
Course material: see here