Academic interests
I work as a scientist at the Norwegian Defence Research Establishment (FFI) whilst studying for a PhD in statistics at UiO, with Nils Lid Hjort (UiO) and Erik Unneberg (FFI) as supervisors. My main interests are in Bayesian nonparametrics and simulation techniques.
Courses taught
Autumn 2023: STK4021 - Applied Bayesian Analysis
Background
2020 - 2021: MSc in Mathematical Sciences, Wadham College, University of Oxford
2014 - 2018: Master of Mathematics, Mansfield College, University of Oxford
Awards
- FFI's Communication Awards 2023 (best scientific publication) for "Inference for Bayesian nonparametric models with binary response data via permutation counting" in Bayesian Analysis
- NTREM 2023: Best presentation
- FFI's Communication Awards 2022 (open category) for the summer course in Bayesian nonparametrics
- Norway Scholarship 2020
Publications
-
Christensen, Dennis & Moen, Per August Jarval
(2023).
Fast implementation of a general importance sampling algorithm for Bayesian nonparametric models with binary responses.
-
Christensen, Dennis & Moen, Per August Jarval
(2023).
perms: Likelihood-free estimation of marginal likelihoods for binary response data in Python and R.
arXiv.org.
ISSN 2331-8422.
doi:
10.48550/arXiv.2309.01536.
Show summary
In Bayesian statistics, the marginal likelihood (ML) is the key ingredient needed for model comparison and model averaging. Unfortunately, estimating MLs accurately is notoriously difficult, especially for models where posterior simulation is not possible. Recently, Christensen (2023) introduced the concept of permutation counting, which can accurately estimate MLs of models for exchangeable binary responses. Such data arise in a multitude of statistical problems, including binary classification, bioassay and sensitivity testing. Permutation counting is entirely likelihood-free and works for any model from which a random sample can be generated, including nonparametric models. Here we present perms, a package implementing permutation counting. As a result of extensive optimisation efforts, perms is computationally efficient and able to handle large data problems. It is available as both an R package and a Python library. A broad gallery of examples illustrating its usage is provided, which includes both standard parametric binary classification and novel applications of nonparametric models, such as changepoint analysis. We also cover the details of the implementation of perms and illustrate its computational speed via a simple simulation study.
-
Christensen, Dennis; Unneberg, Erik; Høyheim, Eirik; Jensen, Tomas Lunde & Hjort, Nils Lid
(2023).
Improved measurements of impact sensitivities of energetic materials.
Show summary
Accurate estimation of impact sensitivity is crucial for safe production, handling, storage and transport of energetic materials. Indeed, molecular characteristics will affect sensitivity, and for solid materials, factors like particle size, lattice defects and morphology also play a role and make reproducibility difficult. As various synthesis and recrystallisation methods may lead to differences in crystal properties, it is important to determine the impact sensitivity whenever an energetic material is prepared. Of particular interest is the median ℎ50, namely the impact energy level at which there is a probability of 50% of an explosion occurring. This value has been shown to correlate with quantum chemical properties of the energetic material in question, providing insight into the underlying causes which govern sensitivity. However, in practical applications, it may be more important to estimate extreme values like the 99% quantile ℎ99. In addition to providing point estimates, we would like to derive confidence intervals to address their uncertainty. In this work, we cover the most common methods for constructing such confidence intervals (the delta method, Fieller’s theorem and the likelihood-ratio test) and compare their performance on sensitivity data via simulations. Our experiments indicate that Fieller’s theorem is the superior method, and we therefore use it to construct confidence intervals for ℎ50 and ℎ99 for cyclotetramethylene- tetranitramine (HMX), using new data. Based on our results, we formulate recommendations for researchers measuring sensitivities of synthesised molecules.
-
(2022).
Nonparametric Bayesian sensitivity testing with optimal design.
View all works in Cristin
Published
Jan. 25, 2022 3:00 PM
- Last modified
Dec. 18, 2023 2:29 PM