Ida Scheel

• Liu, Qinghua; Vitelli, Valeria; Mannino, Carlo; Frigessi, Arnoldo & Scheel, Ida (2026). Pseudo-Mallows for Efficient Probabilistic Preference Learning. International Conference on Machine Learning (ICML). doi: https:/arxiv.org/pdf/2205.13911. Full text in Research Archive
• Parviero, Riccardo; Hellton, Kristoffer H.; Canright, Geoffrey & Scheel, Ida (2026). How missing edges affect estimation and prediction accuracy in the graph-based Poisson process model. International Journal of Information Management Data Insights. 6(2), p. 100429–100429. doi: 10.1016/j.jjimei.2026.100429. Full text in Research Archive Show summary

  Spreading processes in a population, such as the adoption of a novel innovation or product, may be driven by both viral interactions and external factors. The social graph of the population is usually used to model peer-to-peer interactions, but in practice, such graphs are never completely observed. In particular, certain interactions may be missing. We conduct a simulation study to determine how missing interactions affect the inference and predictions of the graph-based Poisson process model introduced by Parviero et al. (2022), modelling both peer-to-peer and external influences. If the social graph is only partly observed, simulations show that parameter estimation will be biased. The external share of the overall cumulative intensity is seen to be overestimated, while the viral share is underestimated. The estimated parameters are seen to increase. Simulations suggest that the bias will be relatively small even for high proportions of missing interactions. In addition, the simulations suggest that resulting predictions are not particularly affected.

• Scheel, Ida & Gåsemyr, Jørund Inge (2026). Graphical and Summary Diagnostics for Node Level Adequacy in Bayesian Hierarchical Models. Bayesian Analysis. ISSN 1936-0975. 21(2), p. 485–510. doi: 10.1214/24-BA1478. Full text in Research Archive Show summary

  For Bayesian hierarchical models represented by directed acyclic graphs, all neighbours of a node provide, potentially conflicting, information on the node. Hence identifying potential conflict between these different information sources can be important for assessing the adequacy of Bayesian hierarchical models. Building on this idea originating from O’Hagan (2003), Gåsemyr and Natvig (2009) constructed node based conflict measures that have been shown to be well calibrated. Sharing the idea of considering potential conflict between separate information sources for each node, Scheel et al. (2011) constructed the graphical diagnostic the local critique plot (lcp). This can be used both to identify conflict at each node, and provide insight into the nature of the conflict, hence being more informative than summary diagnostics. In the present paper we link these two ideas together by constructing a new diagnostic plot, the integrated information contributions local critique plot (iic-lcp), that is supplementary to lcp, but builds on the framework of Gåsemyr and Natvig (2009). It has the advantage over the lcp with the possibility to display curves corresponding to different parameters in the same plot, saving space and easing comparisons, particularly useful for sets of parameters representing exchangeable groups. Notably, we establish a relationship between the geometry of the iic-lcp plot and some of the conflict measures defined in Gåsemyr and Natvig (2009). This indicates how the conflict expressed visually in the graphical plots can be soundly assessed numerically.

• Parviero, Riccardo; Hellton, Kristoffer Herland; Canright, Geoffrey & Scheel, Ida (2024). STAR: Spread of innovations on graph structures with the Susceptible-Tattler-Adopter-Removed model. Stat. 13(2). doi: 10.1002/sta4.671. Full text in Research Archive Show summary

  Adoptions of a new innovation such as a product, service or idea are typically drivenboth by peer-to-peer social interactions and by external influence. Social graphs areusually used to efficiently model the peer-to-peer interactions, where new adoptersinfluence their peers to also adopt the innovation. However, the influence to adoptmay also spread through individuals close to the adopters, known as tattlers, whoonly share information regarding the innovation. We extend an inhomogeneousPoisson process model accounting for both external and peer-to-peer influence toinclude an optional tattling stage, and we term the extension the Susceptible-Tattler-Adopter-Removed (STAR) model. In an extensive simulation study, the proposedmodel is shown to be stable and identifiable and to accurately identify tattling whenpresent. Further, using simulations, we show that both inference and prediction ofthe STAR model are quite robust against missing edges in the social graph, a commonsituation in real-world data. Simulations and theoretical considerations demonstratethat, when edges are missing, the STAR model is able to accurately estimate theshares attributed to the external and internal sources of influence. Furthermore, theSTAR model may be used to improve the inference of the external and viral parame-ters and subsequent predictions even when tattling is not part of the real data-generating mechanism.

• Hjort, Anders Dahl; Scheel, Ida; Sommervoll, Dag Einar & Pensar, Johan (2023). Locally interpretable tree boosting: An application to house price prediction. Decision Support Systems. ISSN 0167-9236. 178. doi: 10.1016/j.dss.2023.114106. Full text in Research Archive Show summary

  We introduce Locally Interpretable Tree Boosting (LitBoost), a tree boosting model tailored to applications where the data comes from several heterogeneous yet known groups with a limited number of observations per group. LitBoost constraints the complexity of a Gradient Boosted Trees model in a way that allows us to express the final model as a set of local Generalized Additive Models, yielding significant interpretability benefits while still maintaining some of the predictive power of a Gradient Boosted Trees model. We use house price prediction as a motivating example and demonstrate the performance of LitBoost on a data set of observations from different city districts in Oslo (Norway). We also test the robustness of LitBoost in an extensive simulation study on a synthetic data set.

• Parviero, Riccardo; Hellton, Kristoffer Herland; Haug, Ola; Engø-Monsen, Kenth; Rognebakke, Hanne Therese Wist & Canright, Geoffrey [Show all 8 contributors for this article] (2022). An agent-based model with social interactions for scalable probabilistic prediction of performance of a new product. International Journal of Information Management Data Insights. 2(2). doi: 10.1016/j.jjimei.2022.100127. Full text in Research Archive Show summary

  Understanding the spreading process of new products provides valuable knowledge that can be used for effective marketing. The ability to make early prediction of success or failure is a great advantage in innovation processes. Extending current literature in a novel way, we propose a data-driven agent-based methodology that accomplishes this task. Inference and predictions are based on short-time observations of the product adoption history and knowledge of the social network of consumers. We model and predict adoptions at the agent level as driven by unobserved peer-to-peer influence and external factors such as marketing. The method compares interaction between consumers and general campaigns, and quantifies the importance of characteristics of customers and their social relations. Our computationally efficient method is demonstrated by analyzing real data, predicting the process far into the future using data from a short period after launch, and validated by simulation experiments on a true full-scale communication network.

• Hjort, Anders Dahl; Pensar, Johan; Scheel, Ida & Sommervoll, Dag Einar (2022). House price prediction with gradient boosted trees under different loss functions. Journal of Property Research. ISSN 0959-9916. 39(4), p. 333–364. doi: 10.1080/09599916.2022.2070525. Full text in Research Archive Show summary

  Many banks and credit institutions are required to assess the value of dwellings in their mortgage portfolio. This valuation often relies on an Automated Valuation Model (AVM). Moreover, these institutions often report the models accuracy by two numbers: The fraction of predictions within ±20% and ±10% range from the true values. Until recently, AVMs tended to be hedonic regression models, but lately machine learning approaches like random forest and gradient boosted trees have been increasingly applied. Both the traditional approaches and the machine learning approaches rely on minimising mean squared prediction error, and not the number of predictions in the ±20% and ±10% range. We investigate whether introducing a loss function closer to the AVMs actual loss measure improves performance in machine learning approaches, specifically for a gradient boosted tree approach. This loss function yields an improvement from 89.4% to 90.0% of predictions within ±20% of the true value on a data set of N=126719 transactions from the Norwegian housing market between 2013 and 2015, with the biggest improvements in performance coming from the lower price segments. We also find that a weighted average of models with different loss functions improves performance further, yielding 90.4% of the observations within ±20% of the true value.

• Kvamme, Håvard; Borgan, Ørnulf & Scheel, Ida (2019). Time-to-Event Prediction with Neural Networks and Cox Regression. Journal of machine learning research. ISSN 1532-4435. 20(129), p. 1–30. Full text in Research Archive Show summary

  New methods for time-to-event prediction are proposed by extending the Cox proportional hazards model with neural networks. Building on methodology from nested case-control studies, we propose a loss function that scales well to large data sets and enables fitting of both proportional and non-proportional extensions of the Cox model. Through simulation studies, the proposed loss function is verified to be a good approximation for the Cox partial log-likelihood. The proposed methodology is compared to existing methodologies on real-world data sets and is found to be highly competitive, typically yielding the best performance in terms of Brier score and binomial log-likelihood. A python package for the proposed methods is available at https://github.com/havakv/pycox.

• Gåsemyr, Jørund Inge & Scheel, Ida (2019). Alternatives to post-processing posterior predictive p-values. Scandinavian Journal of Statistics. ISSN 0303-6898. 46(4), p. 1252–1273. doi: 10.1111/sjos.12393. Full text in Research Archive Show summary

  The posterior predictive p value (ppp) was invented as a Bayesian counterpart to classical p values. The methodology can be applied to discrepancy measures involving both data and parameters and can, hence, be targeted to check for various modeling assumptions. The interpretation can, however, be difficult since the distribution of the ppp value under modeling assumptions varies substantially between cases. A calibration procedure has been suggested, treating the ppp value as a test statistic in a prior predictive test. In this paper, we suggest that a prior predictive test may instead be based on the expected posterior discrepancy, which is somewhat simpler, both conceptually and computationally. Since both these methods require the simulation of a large posterior parameter sample for each of an equally large prior predictive data sample, we furthermore suggest to look for ways to match the given discrepancy by a computation‐saving conflict measure. This approach is also based on simulations but only requires sampling from two different distributions representing two contrasting information sources about a model parameter. The conflict measure methodology is also more flexible in that it handles non‐informative priors without difficulty. We compare the different approaches theoretically in some simple models and in a more complex applied example.

• Liu, Qinghua; Reiner, Andrew Henry; Rattalma, Arnoldo Frigessi Di & Scheel, Ida (2019). Diverse personalized recommendations with uncertainty from implicit preference data with the Bayesian Mallows model. Knowledge-Based Systems. ISSN 0950-7051. p. 1–12. doi: 10.1016/j.knosys.2019.104960. Full text in Research Archive Show summary

  Clicking data, which exists in abundance and contains objective user preference information, is widely used to produce personalized recommendations in web-based applications. Current popular recommendation algorithms, typically based on matrix factorizations, often focus on achieving high accuracy. While achieving good clickthrough rates, diversity of the recommended items is often overlooked. Moreover, most algorithms do not produce interpretable uncertainty quantifications of the recommendations. In this work, we propose the Bayesian Mallows for Clicking Data (BMCD) method, which simultaneously considers accuracy and diversity. BMCD augments clicking data into compatible full ranking vectors by enforcing all the clicked items clicked by a user to be top-ranked regardless of their rarity. User preferences are learned using a Mallows ranking model. Bayesian inference leads to interpretable uncertainties of each individual recommendation, and we also propose a method to make personalized recommendations based on such uncertainties. With a simulation study and a real life data example, we demonstrate that compared to state-of-the-art matrix factorization, BMCD makes personalized recommendations with similar accuracy, while achieving much higher level of diversity, and producing interpretable and actionable uncertainty estimation.

• Liu, Qinghua; Crispino, Marta; Scheel, Ida; Vitelli, Valeria & Rattalma, Arnoldo Frigessi Di (2019). Model-based learning from preference data. Annual Review of Statistics and Its Application. ISSN 2326-8298. 6, p. 329–354. doi: 10.1146/annurev-statistics-031017-100213. Full text in Research Archive
• Bolin, David; Frigessi, Arnoldo; Guttorp, Peter; Haug, Ola; Orskaug, Elisabeth & Scheel, Ida [Show all 7 contributors for this article] (2016). Calibrating regionally downscaled precipitation over Norway through quantile-based approaches. Advances in Statistical Climatology, Meteorology and Oceanography (ASCMO). ISSN 2364-3579. 2, p. 39–47. doi: 10.5194/ascmo-2-39-2016. Full text in Research Archive Show summary

  Dynamical downscaling of earth system models is intended to produce high-resolution climate information at regional to local scales. Current models, while adequate for describing temperature distributions at relatively small scales, struggle when it comes to describing precipitation distributions. In order to better match the distribution of observed precipitation over Norway, we consider approaches to statistical adjustment of the output from a regional climate model when forced with ERA-40 reanalysis boundary conditions. As a second step, we try to correct downscalings of historical climate model runs using these transformations built from downscaled ERA-40 data. Unless such calibrations are successful, it is difficult to argue that scenario-based downscaled climate projections are realistic and useful for decision makers. We study both full quantile calibrations and several different methods that correct individual quantiles separately using random field models. Results based on cross-validation show that while a full quantile calibration is not very effective in this case, one can correct individual quantiles satisfactorily if the spatial structure in the data are accounted for. Interestingly, different methods are favoured depending on whether ERA-40 data or historical climate model runs are adjusted.

• Scheel, Ida; Ferkingstad, Egil; Frigessi, Arnoldo; Haug, Ola; Hinnerichsen, Mikkel & Meze-Hausken, Elisabeth (2013). A Bayesian hierarchical model with spatial variable selection: the effect of weather on insurance claims. Journal of the Royal Statistical Society, Series C (Applied Statistics). ISSN 0035-9254. 62, p. 85–100. doi: 10.1111/j.1467-9876.2012.01039.x. Full text in Research Archive Show summary

  Climate change will affect the insurance industry. We develop a Bayesian hierarchical statistical approach to explain and predict insurance losses due to weather events at a local geographic scale. The number of weather-related insurance claims is modelled by combining generalized linear models with spatially smoothed variable selection. Using Gibbs sampling and reversible jump Markov chain Monte Carlo methods, this model is fitted on daily weather and insurance data from each of the 319 municipalities which constitute southern and central Norway for the period 1997–2006. Precise out-of-sample predictions validate the model. Our results show interesting regional patterns in the effect of different weather covariates. In addition to being useful for insurance pricing, our model can be used for short-term predictions based on weather forecasts and for long-term predictions based on downscaled climate models.

• Scheel, Ida & Hinnerichsen, Mikkel (2012). The Impact of Climate Change on Precipitation-related Insurance Risk: A Study of the Effect of Future Scenarios on Residential Buildings in Norway. Geneva papers on risk and insurance. Issues and practice. ISSN 1018-5895. 37(2), p. 365–376. doi: 10.1057/gpp.2012.7. Full text in Research Archive
• Scheel, Ida; Green, Peter J. & Rougier, Jonathan C. (2011). A Graphical Diagnostic for Identifying Influential Model Choices in Bayesian Hierarchical Models. Scandinavian Journal of Statistics. ISSN 0303-6898. 38(3), p. 529–550. doi: 10.1111/j.1467-9469.2010.00717.x. Full text in Research Archive
• Orskaug, Elisabeth; Scheel, Ida; Frigessi, Arnoldo; Guttorp, Peter; Haugen, Jan Erik & Tveito, Ole Einar [Show all 7 contributors for this article] (2011). Evaluation of a dynamic downscaling of precipitation over the Norwegian mainland. Tellus A: Dynamic Meteorology and Oceanography. ISSN 0280-6495. 63(4), p. 746–756. doi: 10.1111/j.1600-0870.2011.00525.x. Full text in Research Archive
• Parr, CL; Hjartåker, Anette; Hjartaker, A; Scheel, Ida; Lund, Eiliv & Laake, Petter [Show all 8 contributors for this article] (2008). Comparing methods for handling missing values in food-frequency questionnaires and proposing k nearest neighbours imputation: effects on dietary intake in the Norwegian Women and Cancer study (NOWAC). Public Health Nutrition (PHN). ISSN 1368-9800. 11, p. 361–370. doi: 10.1017/S1368980007000365. Full text in Research Archive
• Scheel, Ida; Aldrin, Magne; Frigessi, Arnoldo & Jansen, Peder A (2007). A stochastic model for infectious salmon anemia (ISA) in Atlantic salmon farming. Journal of the Royal Society Interface. ISSN 1742-5689. 4. doi: 10.1098/rsif.2007.0217. Full text in Research Archive Show summary

  Infectious salmon anemia (ISA) is one of the main infectious diseases in Atlantic salmon farming with major economical implications. Despite the strong regulatory interventions, the ISA epidemic is not under control, worldwide. We study the data covering salmon farming in Norway from 2002 to 2005 and propose a stochastic space-time model for the transmission of the virus. We model seaway transmission between farm sites, transmission through shared management and infrastructure, biomass effects and other potential pathways within the farming industry. We find that biomass has an effect on infectiousness, the local contact network and seaway distance of 5 km represent similar risks, but a large component of risk originates from other sources, among which are possibly infected salmon smolt and boat traffic.

• Parr, Christine Louise; Hjartåker, Anette; Scheel, Ida; Lund, Eiliv; Laake, Petter & Veierød, Marit Bragelien (2007). Comparing methods for handling missing values in food-frequency questionnaires and proposing k nearest neighbours imputation: effects on dietary intake in the Norwegian Women and Cancer study (NOWAC). Public Health Nutrition (PHN). ISSN 1368-9800. Full text in Research Archive
• Scheel, Ida; Aldrin, Magne; Glad, Ingrid Kristine; Sorum, Ragnhild; Lyng, Heidi & Frigessi, Arnoldo (2005). The influence of missing value imputation on detection of differentially expressed genes from microarray data. Bioinformatics. ISSN 1367-4803. 21(23), p. 4272–4279. Full text in Research Archive

View all works in NVA

• Aalen, Odd O.; Frigessi, Arnoldo; Moger, Tron Anders; Scheel, Ida; Skovlund, Eva & Veierød, Marit Bragelien (2018). Statistiske metoder i medisin og helsefag. 2. utgave. Gyldendal Akademisk. ISBN 9788205507647. 351 p. Full text in Research Archive
• Aalen, Odd O.; Frigessi, Arnoldo; Moger, Tron Anders; Scheel, Ida; Skovlund, Eva & Veierød, Marit Bragelien (2006). Statistiske metoder i medisin og helsefag. Gyldendal Akademisk. ISBN 9788205346857. 335 p. Full text in Research Archive

View all works in NVA

• Zuccato, Lorenzo; George, Anne-Marie; Vitelli, Valeria; Frigessi, Arnoldo; Giese, Martin & Scheel, Ida (2026). Bridging Statistical and Logical Perspectives on Inconsistency. Full text in Research Archive
• Haug, Ola; Bolin, David; Frigessi, Arnoldo; Guttorp, Peter; Orskaug, Elisabeth & Scheel, Ida [Show all 7 contributors for this article] (2016). Modelling and predicting residential water damage insurance claims via a calibrated dynamical downscaling. Full text in Research Archive
• Haug, Ola; Frigessi, Arnoldo; Scheel, Ida & Guttorp, Peter (2015). Modelling and predicting residential water damage insurance claims in a climate change perspective. Full text in Research Archive
• Haug, Ola; Scheel, Ida; Orskaug, Elisabeth; Frigessi, Arnoldo; Guttorp, Peter & Ferkingstad, Egil (2014). Vulnerability models for water damage insurance claims - predictions of future losses in a climate change perspective. Full text in Research Archive
• Haug, Ola; Orskaug, Elisabeth; Scheel, Ida; Frigessi, Arnoldo; Maraun, Douglas & Guttorp, Peter (2012). Evaluation and calibration of dynamically downscaled precipitation over Norwegian mainland. Full text in Research Archive
• Haug, Ola; Orskaug, Elisabeth; Scheel, Ida; Frigessi, Arnoldo; Maraun, Douglas & Guttorp, Peter (2012). Evaluation and Calibration of Dynamically Downscaled Precipitation over Norwegian Mainland. Full text in Research Archive
• Scheel, Ida; Haug, Ola; Orskaug, Elisabeth; Frigessi, Arnoldo & Guttorp, Peter (2012). Evaluating and Calibrating Dynamically Downscaled Precipitation Using the Doksum Shift Function. Full text in Research Archive
• Orskaug, Elisabeth; Scheel, Ida; Frigessi, Arnoldo; Guttorp, Peter; Haugen, Jan Erik & Tveito, Ole Einar [Show all 7 contributors for this article] (2011). Evaluation of a dynamic downscaling of Norwegian precipitation. Full text in Research Archive
• Haug, Ola; Orskaug, Elisabeth; Scheel, Ida; Frigessi, Arnoldo; Guttorp, Peter & Maraun, Douglas (2011). Calibrating dynamically downscaled precipitation using the Doksum shift function. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2011). A Graphical Diagnostic for Identifying Influential Model Choices in Bayesian Hierarchical Models. Full text in Research Archive
• Haug, Ola; Orskaug, Elisabeth; Scheel, Ida; Frigessi, Arnoldo; Guttorp, Peter & Maraun, Douglas (2011). Calibrating dynamically down-scaled precipitation using the Doksum shift function. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2010). A Graphical Diagnostic for Identifying Influential Model Choices in Bayesian Hierarchical Models. Full text in Research Archive
• Orskaug, Elisabeth; Haug, Ola; Scheel, Ida & Frigessi, Arnoldo (2010). A validation suite for downscaled climate model data. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2010). A Graphical Diagnostic for Identifying Influential Model Choices in Bayesian Hierarchical Models. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2009). A graphical diagnostic for identifying influential model choices in Bayesian hierarchical models. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2008). Identifying influential model choices in Bayesian hierarchical models. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2008). Identifying influential model choices in Bayesian hierarchical models. Full text in Research Archive
• Lyngstad, Trude M.; Scheel, Ida & Jansen, Peder A. (2007). Hvordan spres infeksiøs lakseanemi (ILA)? Norsk Fiskeoppdrett. ISSN 0332-7132. 32(9), p. 52–55. Full text in Research Archive
• Scheel, Ida; Aldrin, Magne; Frigessi, Arnoldo & Jansen, Peder A (2007). A stochastic model for infectious salmon anemia (ISA) in Atlantic salmon farming. Full text in Research Archive
• Parr, Christine Louise; Hjartåker, Anette; Scheel, Ida; Laake, Petter; Lund, Eiliv & Veierød, Marit Bragelien (2006). Substituting missing values in food frequency questionnaires (FFQs): effects on energy intake in the Norwegian Women and Cancer Study (NOWAC). Full text in Research Archive Show summary

  Objective: Missing values are common in FFQs used in large epidemiological studies and must be handled when calculating energy intake. We compare results from different methods for substituting/imputing missing values using NOWAC data. Methods: A FFQ was mailed twice (test-retest) to 1995 women aged 46-75 y from the NOWAC cohort in a reproducibility study (75% response). Missing answers in the test were imputed as follows: 1. Frequencies=0 (null intake) and portion sizes=smallest 2. Mode values 3. Median values 4. Retest values. Remaining missing values were treated as in method 1 5. K-nearest neighbors imputation using a weighted average of the values for the same question from the K=10 most similar respondents within the same dataset. Results: After imputation of the test FFQ (17% missing data) with the different methods, estimated mean energy intake (MJ/day) with 95% CI was: 1. 6.43 (6.34, 6.53) 2. 6.92 (6.83, 7.01) 3. 7.16 (7.06, 7.25) 4. 6.93 (6.84, 7.03) 5. 7.52 (7.42, 7.62) Conclusion: Exclusion is not practical when most respondents have missing values. However, the calculated energy intake is influenced by the imputation method. Treating missing as null intake may not be correct for all foods and may lead to underestimation of energy intake.

• Scheel, Ida; Aldrin, Magne; Glad, Ingrid Kristine; Sørum, R.; Lyng, Heidi & Frigessi, Arnoldo (2006). The influence of missing value imputation on the detection of differentially expressed genes from microarray data. Full text in Research Archive
• Scheel, Ida; Aldrin, Magne; Glad, Ingrid Kristine; Sørum, Ragnhild; Lyng, Heidi & Frigessi, Arnoldo (2005). The influence of missing value imputation on the detection of differentially expressed genes from microarray data. Full text in Research Archive
• Parr, Christine Louise; Hjartåker, Anette; Scheel, Ida; Laake, Petter; Lund, Eiliv & Veierød, Marit Bragelien (2005). Substituting missing values in food frequency questionnaires (FFQs): effects on energy intake in the Norwegian Women and Cancer Study. Full text in Research Archive Show summary

  Objective: Missing values are common in self-administered FFQs used in large epidemiological studies and must be handled when calculating energy and nutrient intake. We compare results from different methods for substituting/imputing missing values using data from the Norwegian Women and Cancer study (NOWAC). We include K-nearest neighbors imputation (KNNimpute). This is a widely used method for missing entries in cDNA microarray data, but is here extended and adapted to FFQ data, which is new to our knowledge. Methods: A FFQ designed to assess habitual diet was mailed twice (test-retest 3 months apart) to a random sample of 1995 women aged 46-75 y from the NOWAC cohort as part of a reproducibility study (75% response). A total of 126 questions were included in the dietary intake calculations. Missing answers to consumption frequencies and portion sizes in the test FFQ were imputed before the calculations using the following methods: 1. Frequencies=0 (null intake) and portion sizes=smallest for a conservative estimate 2. The mode value 3. The median value 4. Non-missing values in the retest. Remaining missing values were treated as in method 1 5. KNNimpute using a weighted average of the values for the same question from the K=10 most similar respondents within the same dataset (test). Similarity is evaluated from the closeness of responses to the other FFQ questions. Results: In the test FFQ 17% of the data matrix was missing. Among respondents (n=1495) 95% had ≥1 missing answer. After imputation with retest values (method 4), 10% of the data was still missing and treated as in method 1. The estimated mean energy intake (MJ/day) with 95% CI after imputation according to the different methods was: 1. 6.43 (6.34, 6.53) for null intake and the smallest portion 2. 6.92 (6.83, 7.01) for mode 3. 7.16 (7.06, 7.25) for median 4. 6.93 (6.84, 7.03) for retest 5. 7.52 (7.42, 7.62) for KNNimpute Conclusion: Exclusion is not practical when most respondents have missing values. However, the calculated energy intake is influenced by the imputation method used. Missing is frequently handled as null intake, but this may not be correct for all foods. This is indicated by the increase in energy when missing is imputed by the other methods, including the retest values from the same respondents 3 months later. KNNimpute gave the highest energy intake, and will be investigated further.

• Scheel, Ida; Aldrin, Magne; Glad, Ingrid Kristine; Sørum, R.; Lyng, Heidi & Frigessi, Arnoldo (2005). The influence of missing value imputation on the detection of differentially expressed genes from microarray data. Full text in Research Archive
• Scheel, Ida (2004). Spread of infectious agents in salmon farming. Full text in Research Archive
• Scheel, Ida; Ferkingstad, Egil; Frigessi, Arnoldo; Haug, Ola; Hinnerichsen, Mikkel & Meze-Hausken, Elisabeth (2011). A Bayesian hierarchical model with spatial variable selection: the effect of weather on insurance claims. Derivation of distributions and MCMC sampling schemes. Universitetet i Oslo. Full text in Research Archive Show summary

  Climate change will affect the insurance industry. We develop a Bayesian hierarchical statistical approach to explain and predict insurance losses due to weather events at a local geographical scale. The number of weather-related insurance claims is modelled combining generalized linear models with spatially smoothed variable selection. Using Gibbs sampling and reversible jump MCMC, the model is fitted on daily weather and insurance data from each of the 319 municipalities of southern and central Norway for the period 1997-2006. Out-of-sample predictions from the model are very good. Our results show interesting regional patterns in the impact of different weather covariates. In addition to being useful for insurance pricing, our model can be used for short-term predictions based on weather forecasts and long-term predictions based on downscaled climate models.

• Scheel, Ida & Hinnerichsen, Mikkel (2010). The Impact of Climate Change on Insurance Risk: A Study of the Effect of Climate Change Scenarios in Norway. Norwegian Computing Center. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2010). Applications of the Local critique plot. Universitetet i Oslo. Full text in Research Archive Show summary

  The use of the Local critique plot is illustrated by three different applications. The applications involve directed acyclic graph (DAG) models of varying complexity and structure, and they illustrate different aspects and abilities of the plots. We see examples where the posterior is a trade off of tail specifications because of conflict between the local prior and the lifted likelihood. Hence model choices made on these tails are, perhaps unexpectedly, identified as influential. We also see how the local critique plot can reveal local prior or lifted likelihood domination. The applications also illustrate that diagnostic warnings from the local critique plot prompt us to reconsider the statistical model, but that sometimes a reexamination can assure us that the choices are justified.

• Orskaug, Elisabeth; Scheel, Ida; Frigessi, Arnoldo; Guttorp, Peter; Haugen, Jan Erik & Tveito, Ole Einar [Show all 7 contributors for this article] (2010). Supplemental material to: Evaluation of a dynamic downscaling of Norwegian precipitation. Norsk Regnesentral. Full text in Research Archive
• Scheel, Ida (2008). Influence of incomplete knowledge in inference: cases in the life sciences. Faculty of Mathematics and Natural Sciences, University of Oslo. Full text in Research Archive
• Scheel, Ida; Green, Peter J & Rougier, Jonathan C (2008). Identifying influential model choices in Bayesian hierarchical models. Department of Mathematics, University of Bristol. Full text in Research Archive
• Frigessi, Arnoldo & Scheel, Ida (2002). A first statistical methods for the geofraphic localisation of GSM mobile phones. Full text in Research Archive

View all works in NVA