Touqeer Ahmad
I propose the following master project, which is about modelling dependence in discrete multivariate extremes: project description.
Riccardo De Bin
My main fields of research are statistical learning (STK-IN4300) and statistical inference (STK4011). Potential topics for a master project include:
- Statistical Boosting: statistical boosting is a method which combines the power of a machine-learning approach with the interpretability of a statistical model. Potential projects could either be methodological or applied.
- Inference in non-regular settings: asymptotic results, such as Wilks' theorem, hold under specific conditions, including the requirement that the true value of the parameter lies in the interior of the parameter space. The latter condition is not met in various situations, including certain experiments in high-energy physics. The goal of a possible thesis project in this area would be to evaluate the methods used in physics and assess their validity.
Examples of master's theses that I have supervised:
Vegard Stikbakke. A boosting algorithm to extend first-hitting-time models to a high-dimensional survival setting. Long thesis, 2019.
Jonas Gjesvik. Statistical modelling of Goalkeepers in the Norwegian Tippeliga. Long thesis, 2019.
Lars H. B. Olsen. Likelihood-Based Boosting: Approximate Confidence Bands and Intervals for Generalized Additive Models. Long thesis, 2020.
Gudmund Horn Hermansen
My primary research areas include time series modeling, the econometrics of high-frequency data, Bayesian statistics, and machine learning. The majority of the master's theses I have supervised have been application-oriented. I am especially interested in applications within high-frequency finance and peace research. I collaborate with the VIEWS research consortium at PRIO, which focuses on developing methods to forecast violent conflicts and their impacts on society and human development.
Examples of master's theses that I have supervised:
Pål Ravlo Hersleth: Conformal Inference for Financial Time Series - Uncertainty Quantification for Volatility Estimates
Jens Helleberg Jakobsen: Application of count time series to battle deaths
Nils Lid Hjort
I am professor emeritus in statistics (office #810, 8th floor), but defined as a sufficiently active one to still be interested in research and supervision. For some of my themes, check out the FocuStat webpage linked to here, with information about earlier projects for Masters and PhDs, publications, talks, blog posts, etc. I have supervised above forty Master and Master-like projects earlier, and if I should take on no. n + 1 or even n + 2, I would do it with a co-supervisor, like Emil Stoltenberg or Ingrid Hobæk Haff (or Ingrid Glad, or Celine Cunen, or Dennis Christensen).
Emil and I are in the process of completing a somewhat Big Book, the second half of which will be 100 Statistical Stories (no less) -- with Sports, History, Music, Social Sciences, Biology, Demography, Simulation, etc. If interested, send us a mail, and we can send you a "current pdf" of these. You could then skim through, find 5-6-7 of these which to a sufficent pedagogical degree intrigue you, and then we can attempt to find out whether associated themes could lead to a project.
Ingrid Hobæk Haff
My main field of research over the past years has been multivariate modelling, in particular copulas, with applications within for instance finance, insurance and climate. I am involved in the Centre of Excellence Integreat, where I among others am working on generation of synthetic data and variational inference, building on copulas. Possible master topics include:
- Synthetic data generation: the setting of interest is one where the true data are sensitive, and protection of privacy must be taken into account when generating the data. Potential projects involve extensions of an existing copula based framework to more complex situations or higher dimensions.
- Vine copula based variational inference: potential projects will build on a vine copula based variational inference method, where the aim would be for instance to extend the methodology and corresponding software to more flexible models, or to apply the existing model to estimate occupancy models, where the aim is to track changes in species abundance over time and space.
Examples of master's theses that I have supervised:
Elisabeth Griesbauer: Vine Copula Based Synthetic Data Generation for Classification. Long thesis, 2022.
Heming Smedsrud Aldrin: Portfolio management with vine copulas. Short thesis, 2023.
Hanne Gjemdal: Fraud Detection Using Vine Copula Regression. Long thesis, 2024.
Berit Omli Øksnes: Fast approximate loss function for vine copulas. Long thesis, 2025.
Aliaksandr Hubin
I am mainly specializing in Bayesian model selection, averaging, and automatic model configuration in complex regression contexts ranging from simple linear models to highly nonlinear Bayesian logic regressions (BLR) [1], Bayesian generalized nonlinear models (BGNLM) [2] and Bayesian neural networks (BNN) [3]. Potential master projects with me could be on further developing variational inference within BGNLM [2]. Alternatively, we can work on further developing subsampling in MCMC [4] for model selection. The last direction is related to BNN [2], here the topics are a. developing interpretable BNN [3] through full structure learning allowing to infer the level of nonlinearity per covariate and the importance of covariates overall; b. developing space state BNN through modeling weights as random effects with a flexible spatio-temporal structure based on normalizing flows.
Examples of some master theses I have supervised or co-supervised:
Lachmann, Jon. "Subsampling Strategies for Bayesian Variable Selection and Model Averaging in GLM and BGNLM." (2021). @Stockholm University, 30 ECTS
Skaaret-Lund, Lars. "Improving latent binary Bayesian neural networks using the local reparametrization trick and normalizing flows." (2022). @University of Oslo, 60 ECTS
Hauglie Sommerfelt, Philip Sebastian. "Combining Variational Bayes and GMJMCMC for Scalable Inference on Bayesian Generalized Nonlinear Models." (2023). @University of Oslo, 60 ECTS
Ellingsen, Herman. "Outlier Detection in Bayesian Neural Networks." (2023). @NMBU, 30 ECTS
References:
- [1] Hubin, Aliaksandr, Geir Storvik, and Florian Frommlet. "A novel algorithmic approach to Bayesian logic regression (with discussion)." Bayesian Analysis (2020)
- [2] Hubin, Aliaksandr, Geir Storvik, and Florian Frommlet. "Flexible Bayesian Nonlinear Model Configuration." Journal of Artificial Intelligence Research (2021).
- [3] Hubin, Aliaksandr, and Geir Storvik. "Variational Inference for Bayesian Neural Networks under Model and Parameter Uncertainty." arXiv:2305.00934 (2023).
- [4] Lachmann, Jon, Geir Storvik, Florian Frommlet, and Aliaksandr Hubin. "A subsampling approach for Bayesian model selection." International Journal of Approximate Reasoning (2022)
Odd Kolbjørnsen
My main fields of research are Bayesian statistics and machine learning, utilizing methods of computational statistics (STK 4051). Phrasing methods from deep learning in the framework of Bayesian statistics provides a rigorous interpretation of knowledge and uncertainty.
One topic of research is inverse problems. Inverse problems arise from indirect measurements, and have applications in medical imaging, remote sensing and for geophysical data. A typical example would be to reconstruct a high resolution image from a low resolution image. The interesting part is that the solution is non-unique, and we will investigate how to use methods from deep learning to explore the space of possible solution, perhaps using methodology from ( Sanchis and Kolbjørnsen, 2023).
Another topic of research is that of Bayesian Neural Nets. The approach of function approximation utilizing random functions (Rahimi and Rech 2008), has recently been studied for deep neural nets (Bolager et al 2023). We want to extend this approach in a setting where weights are trained in a Bayesian framework.
For all master-projects I supervise I recommend courses, STK4021, STK4051, and IN4310 for background, but this is a topic we can discuss.
Examples of master's theses that I have supervised:
Marius Aasan, Invertible and Pseudo-Invertible Encoders: An Approach to Inverse Problems with Neural Networks Long Thesis 2021
Torgeir Ladstein Waagbø, Autoregressive Generative Models with Applications to Super Resolution , Long Thesis 2023
References:
C. Semin-Sanchis and O. Kolbjørnsen, "Sampling-Free Bayesian Inference for Local Refinement in Linear Inversion Problems With a Latent Target Property," in IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1-14, 2023
Ali Rahimi and Benjamin Recht. Uniform approximation of functions with random bases. In 2008 46th Annual Allerton Conference on Communication, Control, and Computing, pages 555–561, Monticello, IL, USA, September 2008. IEEE. ISBN 978-1-4244-2925-7.
Erik Lien Bolager, Iryna Burak, Chinmay V. Datar, Qing Sun, and Felix Dietrich, 2023, Sampling weights of deep neural networks ArXiv: abs/2306.16830
Sven Ove Samuelsen
My field is event history analysis (cf. STK4080) and in particular I have been interested in theoretical developments and applications in epidemiology. Some of the master’s theses I have supervised have had a focus on developments of case-control studies and other epidemiological designs with a time to event perspective, this then includes development of statistical analysis to use for such designs. Others master’s theses have been more directly connected to analyses of specific epidemiological data sets. The problems and data sets for the master’s theses often turn up in connection with collaborations with researchers at the Norwegian Institute of Public Health where I work part time or Norwegian Cancer Registry where I have several collaborators.
Examples of master's theses that I have supervised:
Marthe Kittilsviken Tendrup: Two-phase valdidation correction: A simulation study for robust variance estimation. Long thesis 2022.
Graeme Lawrence: Bacteraemia and Cardiovascular Disease. Long thesis 2020.
Morten Madshus: A Match Too Much? - A simulation study on overmatching in nested case-control studies. Long thesis 2019
Simon Lergenmuller: Two-stage predictor substitution for time-to-event data. Long thesis, 2017.
Ida Scheel
My main fields of research are within various themes within Bayesian statistics. The last years my main research has evolved around preference learning, including how to handle inconsistent preferences. This is research done within the center of excellence in Knowledge-driven Machine Learning Integreat. I am also involved in the interdisciplinary UiO convergence environment BioM, which aims at developing new interdisciplinary methods to model and govern biodiversity under uncertainty, combining ecology, statistics, and philosophy. I am also interested in research on identifying conflict in Bayesian hierarchical models.
Emil Aas Stoltenberg
I am an associate professor in statistics (office number 817, just by the kitchen). I hesitate to clearly define my interests in statistics, as I am sure that you (the master student) can make me want to supervise almost anything in our field, and I am happy to learn new things. So feel free to drop by my office for a non-binding talk!
That being said, there are things I know more about than others, these include survival analysis and high-frequency data (both rely heavily on martingale theory), large sample theory, Bayes and decision theory, and lately some time series analysis. Inside these broad categories, I have a few ongoing projects (papers to be) as well as some more vague ideas, where it would be very nice to get a master student onboard.
Geir Storvik
My main fields of research are (Bayesian) computational statistics and Monte Carlo methods. Sequential Monte Carlo methods in which information needs to be updated as new data is arriving is of particular interest (a method that was used for estimation of the Norwegian reproduction rate for the Covid-19 pandemic). I am also interested in Bayesian versions of Neural networks type regression models.
I am involved in INTEGREAT, a center of excellence in Knowledge-driven Machine Learning.
For all master-projects I supervise, STK4021, STK4051 are useful background courses.
Possible master-project:
- Sequential Monte Carlo for high-dimensional settings.
Examples of master's theses I have supervised:
Martin Strøm Olsen: Twisting targets in Sequential Monte Carlo. Long thesis, 2022
Lars Einar Skaaret-Lund: Improving latent binary Bayesian neural networks using the local reparametrization trick and normalizing flows. Long thesis, 2022
Fredrik Lundvall Wollbraaten: Parallel subsampling MCMC and the Perturbed Subset Parameter Approximation. Long thesis, 2020.
Thordis Thorarinsdottir
My research concerns modelling of randomness and quantification of uncertainty in natural (environmental) processes, particularly spatial and space-time modelling as well as modelling of extremes. The models are commonly constructed with a hierarchical structure where Bayesian inference is a useful way of estimating the model, or hybrid statistical/machine learning approaches.
Since the aim of the modelling is often to issue a prediction, or a forecast, my work also concerns probabilistic forecasting and methods for forecast evaluation, which is highly related to the use of loss functions in machine learning.
Modelling of natural processes requires a combination of statistical theory, the knowledge of statistical and machine learning models, and environmental process understanding. To ensure the last, my work is often conducted in collaboration with scientists from fields such as meteorology, hydrology and climate science.
Some potential master thesis projects are listed here.
Useful background courses are STK4150, STK4021 and STK4051.
Valeria Vitelli
I am a professor at the Department of Biostatistics of UiO, which is part of the Oslo Centre for Biostatistics and Epidemiology (joining statisticians from UiO and OUS). My main fields of research fall within high-dimensional statistics, Bayesian methods, machine learning, and functional data analysis. In general, I have a strong focus on modelling intrinsically smooth processes and complex data structures, often motivated by biomedical applications. I am also a Principal Investigator at Integreat (the Norwegian Centre for Knowledge-driven Machine Learning).
Potential topics for a master project include:
- Multivariate Latent Function-on-Scalar Regression for Anaesthesia Data. This project focuses on extending a novel latent Function-on-Scalar (F-o-S) regression framework designed for observed sequences of correlated binary data [1].
The model extension is motivated by the analysis of the Safe ANesthesia DAtabase (SANDA), where the aim is to evaluate how anaesthetic techniques and dosages influence depth of anaesthesia across patient subgroups by considering multiple sedation proxies simultaneously. Therefore, the methodological aim is to adapt the framework to handle multivariate curves, for instance, by embedding multivariate functional principal component analysis in the current FoS framework. - Latent Function-on-Function Regression. The method described in [1] is a function-on-scalar regression, linking scalar covariates to functional responses. This second project aims at extending the latent functional regression framework to a function-on-function (F-o-F) setup, where both the predictor and the response are functions or continuous time-series. A key application involves using physiological signals (such as multi-channel ECG traces) to model and predict dynamic clinical outcomes over time (such as dissociative episodes in PTSD patients), which is part of a collaboration with the Department of Behavioural Medicine at the Institute of Basic Medical Sciences (see [2,3]).
Both topics fall within the general area of Functional Data Analysis (FDA). If you are interested and to get an idea of what FDA methods entail, you can have a look at my PhD course on FDA here: https://valeriavitelli.github.io/PhDcourse_FDA/
[1] Asgari, F., Alamatsaz, M. H., Hayati, S., & Vitelli, V. (2025). Likelihood analysis of latent functional response regression models for sequences of correlated binary data. Scandinavian Journal of Statistics, 52(2), 840–872. https://doi.org/10.1111/sjos.12773
[2] Gagliano, A., Di Maria, C., Sottile, G., Beutler-Traktovenko, S., Augugliaro, L., & Vitelli, V. (2025, May). A Novel Spectral Density Operator Approach to Unveil Dynamic Time Dependencies in Multivariate Long-Term ECGs. In International Workshop on Functional and Operatorial Statistics (pp. 217-224). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-92383-8_27
[3] Asgari, F., Vitelli, V., & Sailer, U. (2026). Functional Structural Equation Modeling with Latent Variables. Structural Equation Modeling: A Multidisciplinary Journal, 1–9. https://doi.org/10.1080/10705511.2026.2702102
Manuela Zucknick
Cross-Leverage Screening for Bayesian regression in high-dimensional data
Motivation
High-dimensional regression problems, particularly in genomics, involve a very large number of predictors relative to the sample size. Bayesian regression (Gelman et al., 2026) provides a flexible framework for modelling such data and incorporating uncertainty, but computational challenges become substantial as the number of predictors increases. The inclusion of pairwise interactions further expands the model space from O(p) to O(p2). Likewise, in multi-response regression settings the model space expands to O(pm), where m is the number of response variables (Zhao et al., 2021; Zhao et al., 2024). This thesis project proposes to investigate cross-leverage scores (CLS) (Parry et al., 2021; Teschke et al., 2024) as a scalable pre-screening approach for Bayesian regression. Unlike conventional marginal screening, CLS may better identify predictors that are jointly relevant with the response and remain informative in the presence of strong correlations between predictors.
Research Questions
1. Can CLS improve the computational efficiency of high-dimensional Bayesian regression while retaining predictive and inferential performance?
2. How reliably can CLS identify relevant predictors and interactions under strong correlation and weak or non-marginal signals?
3. How does screening affect Bayesian posterior inference and uncertainty quantification?
Proposed Approach
The research will develop a two-stage framework. First, cross-leverage scores will be used to screen a highdimensional predictor space and identify a substantially smaller set of candidate variables. Scalable iterative and randomized methods will be investigated to perform this screening without explicitly constructing large matrices. Second, Bayesian regression models will be fitted to the screened variables. The study will consider Bayesian regression models with shrinkage priors, such as the generalised horseshoe and R2-D2, or alternatively spike-and-slab priors for variable selection. The effect of screening on estimation, variable selection, prediction, and posterior uncertainty will be investigated. Sample splitting will also be considered to assess its potential to reduce post-screening bias.
The proposed framework will be evaluated through simulations covering different sample sizes, dimensionalities, correlation structures and signal strengths. Performance will be assessed using screening accuracy, estimation and prediction error, posterior coverage, runtime, and memory usage. The methodology will subsequently be applied to a large-scale genomic dataset, such as UK Biobank, to assess its practical scalability and ability to identify genetic and gene–environment interactions. Alternatively, if the candidate is interested in exploring the multi-response setting, the methodology would be applied to suitable data sets from large-scale cancer drug screens. The expected contribution is a general, computationally efficient screening framework for high-dimensional Bayesian regression. The thesis will investigate the trade-off between reducing computational complexity and preserving the statistical and inferential properties of Bayesian regression.
References
Teschke, Sven, Katja Ickstadt, and Alexander Munteanu. ”Detecting Interactions in High-Dimensional Data Using Cross Leverage Scores.” Biometrical Journal 66.8 (2024): e70014.
Parry, Katharina, Leo N. Geppert, Alexander Munteanu, and Katja Ickstadt. ”Cross-leverage scores for selecting subsets of explanatory variables.” arXiv preprint arXiv:2109.08399 (2021).
Gelman, Vehtari, McElreath, Simpson, Margossian, Yao, Kennedy, Gabry, Bürkner, Modrák, Leos Barajas (2026). Bayesian Workflow. Chapman Hall.
Zhao, Zhi, Marco Banterle, Leonardo Bottolo, Sylvia Richardson, Alex Lewin, and Manuela Zucknick. "BayesSUR: An R package for high-dimensional multivariate Bayesian variable and covariance selection in linear regression."Journal of Statistical Software100 (2021): 1-32.
Zhao, Zhi, Marco Banterle, Alex Lewin, and Manuela Zucknick. "Multivariate Bayesian structured variable selection for pharmacogenomic studies."Journal of the Royal Statistical Society Series C: Applied Statistics73, no. 2 (2024): 420-443.
Clinical Trials Research Group
We are a research group of statisticians at Oslo Centre for Biostatistics and Epidemiology (OCBE), Oslo University Hospital. We are involved in the practical conduct of randomised controlled trials (RCTs), while also doing research on methodological issues relating to the design, analysis and interpretation of data from such trials. Randomized controlled trials are sometimes described as the gold standard of scientific evidence, but poor design, implementation and adherence can easily invalidate the results. In cases like these, suitable statistical methods can make all the difference! We seek motivated master students for a broad array of topics. The choice of main supervisor will depend on the choice of topic.
Potential supervisors:
Potential projects (just a few examples):
- Defining a continuous measure of compliance: in some trials certain patients will only partially adhere to their randomised treatment (for example they will take their drug on some days, but not every day). How should such patients be included in the analysis? And what is the estimand? See this paper for example.
- Missingness in Markov models: transition models are becoming common in RCTs, but how can we deal with various missingness patterns in such longitudinal data
- RCTs during epidemics: running RCTs during infectious disease outbreaks pose challenges, particularly regarding the causal interpretation of the results. How can issues of contamination and interference be avoided?
- Preferential design: sometimes you would like to include a group of patients in the study which cannot be randomised (maybe they all need to get one of the treatments because of some underlying condition). How can we combine the evidence from the randomized and the non-randomised patients? How would we make inference on the non-randomised population? What assumptions are needed to do so?
- Adapting the sample size using the elicited distribution of the smallest worthwhile effect: The target difference drives a trial's sample size, and for patient-reported outcomes patients are the stakeholders best placed to judge which differences are worth having. They are rarely asked, and when they are, it is in separate pre-trial studies of populations that may differ from the trial's. Imagine asking every screened participant, before randomisation, for their smallest worthwhile effect (SWE): the additional average improvement over the comparator that would make the intervention worth its burdens. Once the first m participants have answered, re-estimate the sample size using a pre-specified quantile of the elicited SWE distribution as the target difference. This adaptation uses only pre-randomisation, non-comparative data, so the type I error rate is unaffected. The target has a direct interpretation: the smallest average effect that a pre-specified share q of patients would judge worthwhile. The price is a random sample size. Derive theory and a full simulation study, conservative and Bayesian rules, and investigate transportability of the elicited distribution.
Contact cmlcunen@uio.no if you have any questions!
Some master theses CC has supervised (with Nils Lid Hjort):
- Marthe Aastveit (2021). Multi-State Models for Interval-Censored Data with Transition Times from Gamma Processes.
- Even Tronstad (2025). Window of Change.