2026-2027

Title : The K-function for point patterns on surfaces

Abtract : The K-function is a well-known summary statistic used to determine whether a point pattern exhibits aggregation, repulsion, or complete spatial randomness. Its development has been applied to different surfaces by leveraging the invariance properties of the surface on which the point pattern is defined, such as the unit sphere or the unit sphere cross time. In this work, we present two extensions of the K-function to more general surfaces. The first extension is based on tangent plane approximations of the surface, while the second relies on geodesic distance. We provide computational algorithms and conduct simulation studies to demonstrate the performance of the extended K-function in detecting various point pattern properties.

Coauthors : Rasmus Waagepetersen (Aalborg University), Edward Cohen (Imperial College London), and Scott Ward (Imperial College London)

Title : Selective inference and its applications

Abtract : Classical statistics centers around the setting where the questions that we ask about our data are specified in advance, before we even look at our data. However, in contemporary settings, often the data analyst comes up with a question only after looking at the data. Classical methods break down in this setting: for instance, standard hypothesis tests do not achieve type 1 error control, and confidence intervals do not attain the nominal coverage. In this brief talk, I will introduce the field of selective inference, which encompasses a number of recent approaches to conduct inference on hypotheses or parameters that are selected from the data, rather than specified in advance.

Title : Minority representation and fairness in network ranking, with an application to school contact diary data

Abstract : Considerations of bias, fairness and representation are a prerequisite of responsible modern statistics, including statistical network analysis. Systematic bias can arise in network analysis when observation errors depend on node attributes, potentially distorting node rankings and group representation. In this talk, we are motivated by a high school contact network constructed from self-reported contact diaries, where nodes represent students, directed edges represent reported contacts, and groups are defined by sex. For networks with two known groups of nodes, we define minority representation as the proportion of nodes from the smaller group among the top-ranked nodes. We characterize its asymptotic behavior in degree-based rankings under a two-group stochastic block model. We model observation error through group-dependent missing-edge mechanisms and develop statistical methods to estimate the error rates and test for systematic bias. When correction is necessary, we apply a ranking correction procedure that targets the estimated group representation expected in the absence of observation error. Application to the contact diary data provides evidence that male students are more likely to omit contacts with female students than vice versa. These findings highlight the importance of accounting for group-dependent observation errors when assessing and correcting group representation in network rankings.  Joint work with Hui Shen and Peter MacDonald.

Titre : Generalized Functional ANOVA: A Complete Theoretical Framework

Résumé : The functional ANOVA provides a fundamental representation of square-integrable multivariate functions into main effects and higher-order interactions. For independent inputs, the components belong to mutually orthogonal Hilbert subspaces and admit an explicit representation. For dependent inputs, however, the components are only hierarchically orthogonal: although existence and uniqueness results are available, the Hilbert subspaces underlying the generalized decomposition have remained implicit. We resolve this representation problem for continuous inputs supported on a bounded hyperrectangle whose joint density is bounded above and away from zero. We introduce a distribution-adapted family of functions and prove that it forms a Riesz basis of the L2 space, thereby guaranteeing a unique, stable, and unconditionally convergent representation. We then show that, for every coalition of variables, the corresponding block of this basis exactly characterizes the Hilbert subspace containing the functional ANOVA component. Our construction recovers the classical orthogonal decomposition under input independence. As a direct consequence, computing the generalized functional ANOVA reduces to estimating coefficients in an explicit, distribution-adapted basis. Finally, as a proof of concept, we introduce an elementary, fast and model-agnostic estimator based on our theoretical results. Experiments on synthetic and real-world datasets illustrate its connections with established tabular explanation methods and show that low order components often capture most of the signal in the model output.

Title : How Many Fish Are in the Sea? Estimating Abundance with Close-Kin Mark-Recapture

Abstract : Knowing how many fish are in the sea is central to fisheries management and conservation, yet abundance is notoriously hard to estimate. Close-kin mark-recapture (CKMR) offers a powerful alternative to traditional surveys and tagging: every offspring genetically “marks” its two parents, so finding parent–offspring and half-sibling pairs among thousands of genotyped samples plays the role of recaptures in classical mark-recapture. Fish need to be sampled only once, samples can come from the catch, and every new fish is compared with all the others, so precision improves quickly as more fish are sampled. Behind this simple idea lies a rich statistical problem. I will describe how CKMR models are built as pairwise pseudo-likelihoods, with kinship probabilities derived from the expected relative reproductive output of each potential parent and linked to an age- and sex-structured population dynamics model. I will discuss how to handle missing or uncertain covariates such as age, and how to tell when models are biased. I will illustrate these ideas with our ongoing study of Atlantic halibut in the Northwest Atlantic, which involves more than 12,000 genotyped fish, incomplete age data, mitochondrial DNA and epigenetic aging clocks. I will close with open problems where mathematics and statistics can make a difference, from inference for dependent pairs to spatial and optimal-design questions.

Titre : Erreur de mesure et modélisation dose-réponse flexible: Une collaboration internationale de comparaisons neutres par simulations.

Résumé : Les erreurs dans la mesure de variables explicatives continues sont un problème fréquent en recherche, dont les conséquences deviennent particulièrement complexes lorsque la relation entre ces variables et la variable réponse est non linéaire. Ce séminaire présentera deux projets interreliés issus d’une collaboration internationale entre deux groupes thématiques de l’initiative STRATOS (STRengthening Analytical Thinking for Observational Studies), consacrés à la modélisation flexible et à l’erreur de mesure. Nous examinerons d’abord, à l’aide d’une étude de simulation, comment l’erreur de mesure peut déformer les relations dose-réponse estimées par splines et polynômes fractionnaires, notamment en masquant certaines non-linéarités ou en créant des courbures artificielles. Nous présenterons ensuite une vaste étude de comparaison neutre par simulation, conçue pour évaluer différentes approches de correction de l’erreur de mesure combinées à des méthodes de modélisation flexible. Une attention particulière sera accordée à son protocole novateur, reposant sur des équipes indépendantes et une évaluation en aveugle, ainsi qu’aux performances comparatives des méthodes SIMEX, bayésiennes, de calibration par régression et d’imputation multiple. Les résultats permettront de dégager plusieurs enseignements méthodologiques pour l’analyse des relations dose-réponse en présence d’erreur de mesure.

Titre : FairGGPR: A multi-criteria fair Gaussian regressor for insurance pricing

Résumé : We study how multiple notions of fairness can be incorporated into a single Bayesian non-parametric regression framework for insurance pricing, with a focus on claim frequency modelling under a log-link. We consider a Generalized Gaussian Process Regression (GGPR) model for count data with risk exposure and introduce fairness interventions in its architecture. Specifically, we address notions of individual fairness by altering the kernel structure to control the similarity between policies (e.g., to mitigate omitted variable bias). We also address group-level fairness by enforcing demographic parity through constraints affecting the posterior. This modified GGPR architecture allows us to jointly enforce multiple fairness notions within a single probabilistic model. We empirically explore trade-offs with actuarial fairness, and how different fairness criteria interact when combined. The results highlight the importance of adopting a multi-criteria, context-aware approach to fairness in insurance pricing.

Titre : Estimation de densité pour données multivariées corrélées et détection de données aberrantes

Résumé : Dans cet exposé, nous aborderons deux aspects des « polygones de fréquences multivariés » (Linear Blend Frequency Polygons, LBFP) en tant qu’estimateurs de densité. Nous commencerons tout d’abord par nommer certaines de leurs propriétés asymptotiques sous des conditions faibles quant à la dépendance entre les observations. Ces propriétés incluent la convergence en moyenne quadratique, les convergences uniformes forte et faible, la convergence L1 et la normalité asymptotique. Quelques considérations pratiques, telles que la réduction du biais et le choix des paramètres de lissage, seront aussi abordées. Ensuite, nous verrons comment, grâce à une astuce computationnelle, ces estimateurs peuvent être utilisés pour la détection de données aberrantes. Des résultats de simulations et d’analyses de données réelles seront présentés dans les deux volets de la présentation, soit l’estimation de densité et la détection de données aberrantes. Ces travaux ont été réalisés en collaboration avec Michel Carbon et Aurélien Nicosia.

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :

Titre :

Résumé :