Research program
How can we obtain reliable inference when data structure is partly hidden, constrained by geometry or sensitive to numerical approximation? I develop model diagnostics and uncertainty quantification methods to address this question. My program connects probabilistic modelling, simulation, visualization and reproducible R software.
Contributions and research themes
Sequential models and latent states
My work with Thierry Duchesne, Louis-Paul Rivest and Daniel Fortin developed multi-state models for animal movement, published in CSDA in 2017 and AOAS in 2017.
In Beyond the next step (2026), I propose generative validation for step-selection functions: comparing simulated trajectories with the movement features that motivate the analysis. My preprints study sequential predictive diagnostics and uncertainty in latent-state sequences.
In Extra hidden states under emission misspecification in hidden Markov models (2026), published in Statistica Neerlandica, I study how additional states can compensate for restrictive emission distributions. A genomic application illustrates the distinction between the number of fitted states and the number of scientifically distinct regimes.
Directional statistics
With Louis-Paul Rivest, Thierry Duchesne and Daniel Fortin, I contributed to a general angular regression model (2016). My 2026 preprint, revised September 30, studies identifiability and likelihood estimation for homogeneous angular regression with correlated random coefficients and a circular random intercept. It distinguishes identification results from the calibration limitations observed in simulations.
In Visualizing directional dependence on cylinders and tori (2026), published in the Journal of Data Science, Statistics, and Visualisation, I propose a conditional graphical grammar for dependence on cylinders and tori. The displays are implemented in ggcircular.
I also develop regression and visualization software connecting the geometry of circular, axial and toroidal data to the statistical questions being studied.
Density and outlier diagnostics
With Thierry Duchesne and Michel Carbon, I develop a leave-one-out influence diagnostic for outlier detection based on linear-blend frequency polygons.
The GLBFP package provides the density estimation software base. Leave-one-out diagnostics are available in its GitHub development version. The presentation of this work received the 2026 SSC New Investigator Presentation Award.
Future directions
The next stages of my program address three questions:
- Localizing failures in sequential models, representing multiple plausible latent trajectories and quantifying error introduced by Monte Carlo approximation.
- Developing diagnostics for hierarchical circular models and dependence tests adapted to circular and toroidal spaces.
- Studying the stability of density diagnostics across dimensions, tuning choices and sampling regimes, with reproducible comparisons of computational cost.
Animal movement is an established application area. Environmental sequences and sensor data are prospective extensions to be developed with domain specialists.
Research training
I favour projects that begin with a focused question, a literature review and reproduction of existing results. Contributions can address theory, simulation, visualization or software validation. Each project should distinguish mathematical results, numerical evidence and open questions.
My experience includes undergraduate project advising, industry internship supervision and examination of master’s theses. Proposed graduate projects build on the three themes above. Teaching and supervision.
Doctoral thesis
This thesis develops a general framework for multi-state modeling of animal movement. It combines the rigor of hidden Markov models with the flexibility of habitat-selection functions, through a unifying formulation connecting behavioral and spatial approaches.
Software and dissemination
R implementations associated with my work are presented in the package catalogue, with their features, documentation and release status.
Current research materials
Code repositories, analyses and prototypes are collected in Open projects. Each entry links to the available materials.
Presentations
- 2026. A Leave-One-Out Influence Statistic for Scalable Density-Based Outlier Detection, Statistical Society of Canada meeting, with Thierry Duchesne and Michel Carbon. Presentation PDF and GitHub repository. This presentation received the 2026 New Investigator Presentation Award.
- 2016. Multi-state Models for Animal Movement Analysis, Statistical Society of Canada meeting, St. Catharines. Original presentation (PDF).
- 2015. Statistique directionnelle et modèle caché… ou comment modéliser le mouvement d’un animal?, statistics seminar, Université Laval, with Thierry Duchesne, Louis-Paul Rivest and Daniel Fortin. Original presentation in French (PDF).
- 2015. Un modèle général de marche aléatoire directionnelle : application au mouvement animalier, ACFAS congress, with Thierry Duchesne and Louis-Paul Rivest. Original presentation in French (PDF).
- 2015. The same theme was presented at the 18th pan-Quebec student colloquium of the Institut des sciences mathématiques. Original presentation in French (PDF).
- 2015. Posters on directional random walks at the SSC meeting and the 20th anniversary of the Association des statisticiennes et statisticiens du Québec.
- 2014. Aventure statistique sur le cercle, with Jessica Bach, Université Laval. Original presentation in French (PDF).
Publications
References updated October 20, 2026. Preprints are listed separately from published journal articles.
Peer-reviewed articles
Nicosia, A. (2026). Extra hidden states under emission misspecification in hidden Markov models. Statistica Neerlandica, 80(4), e70042. https://doi.org/10.1111/stan.70042
Amicizia, D., Marchini, F., Schenone, I., Massaro, E., Landa, P., Nicosia, A., Orsi, A., Icardi, G., Ansaldi, F. (2026). Adjuvanted Influenza Vaccines Versus Other Influenza Vaccines in Adults Aged ≥60 Years: A Systematic Review and Meta-Analysis of Relative Effectiveness for Preventing Influenza-Related Hospitalization. Vaccines, 14(10), 883. https://doi.org/10.3390/vaccines14100883
Nicosia, A. (2026). Visualizing directional dependence on cylinders and tori. Journal of Data Science, Statistics, and Visualisation, 6(7). https://doi.org/10.52933/jdssv.v6i7.181
Nicosia, A. (2026). Beyond the next step: A multi-criteria generative validation framework for step selection functions. Methods in Ecology and Evolution, 17(6), 1754-1767. https://doi.org/10.1111/2041-210X.70313
Gagnon, S., Allard, M., Nicosia, A. (2018). Diurnal and seasonal variations of tundra CO2 emissions in a polygonal peatland near Salluit, Nunavik, Canada. Arctic Science, 4(1), 1-15. https://doi.org/10.1139/AS-2016-0045
Nicosia, A., Duchesne, T., Rivest, L.-P., Fortin, D. (2017). A Multi-State Conditional Logistic Regression Model for the Analysis of Animal Movement. Annals of Applied Statistics, 11(3), 1537-1560. https://doi.org/10.1214/17-AOAS1045
Nicosia, A., Duchesne, T., Rivest, L.-P., Fortin, D. (2017). A General Hidden State Random Walk Model for Animal Movement. Computational Statistics & Data Analysis, 105, 76-95. https://doi.org/10.1016/j.csda.2016.07.009
Rivest, L.-P., Duchesne, T., Nicosia, A., Fortin, D. (2016). A General Angular Regression Model for the Analysis of Data on Animal Movement in Ecology. Journal of the Royal Statistical Society Series C: Applied Statistics, 65(3), 445-463. https://doi.org/10.1111/rssc.12124
Published discussion in statistics education
- Nicosia, A. (2026). Discussion of “Addressing the Challenges of AI-Generated Assignment Submissions in Education: Insights and Strategies”. Journal of Data Science, 24(1), 276-281. https://doi.org/10.6339/26-JDS1208H
Preprints
Nicosia, A., Duchesne, T., Carbon, M. (2026). A Leave-One-Out Influence Statistic for Density-Based Outlier Detection. arXiv:2607.14335. doi:10.48550/arXiv.2607.14335.
Nicosia, A. (2026). Identifiability and likelihood estimation for homogeneous angular regression with random coefficients. arXiv:2607.06790. doi:10.48550/arXiv.2607.06790.
Nicosia, A. (2026). Tropical Viterbi Tubes for Decoding Uncertainty in Hidden Markov Models. arXiv:2606.17181. doi:10.48550/arXiv.2606.17181.
Nicosia, A. (2026). Sequential predictive e-diagnostics for hidden Markov models of animal movement. bioRxiv. doi:10.64898/2026.07.07.737005.
Bouderbala, I., Nicosia, A., Fortin, D. (2026). Behavioural state inference from movement and environmental data using Markovian step selection functions. bioRxiv. doi:10.64898/2026.02.05.704063.