Dr. Beniamino Hadj-Amar

I am a tenure-track Assistant Professor in the Department of Biostatistics at the University of South Carolina (Columbia, SC).

My research lies at the interface of statistics and science, with a particular focus on the development of Bayesian methodologies for the automated analysis of complex dynamical time series.

Prior to joining USC, I was a Postdoctoral Fellow in the Department of Statistics at Rice University, where I worked with Marina Vannucci in collaboration with Read Montague's lab at Virginia Tech. I received my Ph.D. through the Oxford-Warwick Statistics Programme (OxWaSP), under the supervision of Bärbel Finkenstädt at the University of Warwick (UK).

e-mail: hadjamar@mailbox.sc.edu

Research Interests:

Time Series
Hidden Markov Models
Bayesian Inference
Statistical Spectral Analysis
Bayesian Nonparametrics
Variable Selection
Graphical Models
Gaussian Processes

To identify latent structures, deal with complex streams of temporal data, as well as properly account for model uncertainty, I often need to design highly flexible and interpretable models characterized by non-stationary behavior, non-linear features, and sparse structures. From a methodological perspective, I have experience working on switching models (e.g.hidden Markov and semi-Markov models), change-point models, Gaussian processes, Bayesian nonparametrics, graphical models, Bayesian variable selection, factor models, mixture models, statistical spectral analysis, and Markov chain Monte Carlo.

With my collaborators, I have been working on addressing scientific problems across several domains, with a particular focus on data from wearable devices—such as actigraphy and circadian monitoring in patients with epilepsy. I also work extensively in neuroscience, contributing to studies on neuromodulator dynamics using electrophysiological, electrochemical, and fMRI data from the conscious human brain. In addition, I have collaborated on projects in respiratory research, including the modeling of airflow trace data to identify patterns associated with sleep apnea.

Publications:

Batten, S.*, Hadj-Amar, B.*, Ahrens, M.L., Oster, M.E., Hartle, A., Barbosa, L.S., Lohrenz, T., White, J.P., Witcher, M.R., Vannucci, M., Montague, R., Howe, M. & Difeliceantonio, A. Task evoked monoamine release in the human amygdala encodes liking (2025+, submitted).
Hadj-Amar, B., Bornstein, A., Guindani, M., & Vannucci, M., Discrete Autoregressive Switching Processes with Cumulative Shrinkage Priors for Graphical Modeling of Multivariate Time Series Data, Journal of Computational and Graphical Statistics, 2025, (in press).
Hadj-Amar, B., Krishnan, V., & Vannucci, M., Bayesian Covariate-Dependent Circadian Modeling of Rest-Activity Rhythms in Patients with Epilepsy 2025, Data Science in Science, 4 (1).
Batten, S.*, Hartle, A.*, Barbosa. L*, Hadj-Amar, B.*, Bang. D.*, ...,Vannucci, M., & Montague, R., Emotional Words Evoke Region and Valence-Specific Patterns of Concurrent Neuromodulator Release in Human Thalamus and Cortex, 2025, Cell Reports 44(1), 115162.
Hadj-Amar, B., Jewson, J. & Vannucci, M., Bayesian Sparse Vector Autoregressive Switching Models with Application to Human Gesture Phase Segmentation, Annals of Applied Statistics, 2024, 18(3), 2511-2531
Pluta, D*., Hadj-Amar, B.*, Li, M., Zhao, Y., Versace, F. & Vannucci, M., Improved Data Quality and Statistical Power of Trial-Level Event-Related Potentials with Bayesian Random-Shift Gaussian Processes, Nature Portoflio, Scientific Reports, 2024, 14, 8856.
Bang, D.*, Luo, Y.*, Barbosa, L*., Batten, S*., Hadj-Amar, B.*, Twomey, T., Melville, N., White, J., Torres, A., Celaya, X., Ramaiah, P., McClure, S.M., Brewer, G.A., Bina, R.W., Lohrenz, T., King-Casas, B., Chiu, P., Vannucci, M., Kishida, K., Witcher, M. and Montague, P.R, Noradrenaline tracks emotional modulation of attention in human amygdala. Current Biology, 2023, 33 (22) : 5003-5010.
Hadj-Amar, B., Jewson, J. & Fiecas, M., Bayesian Approximations to Hidden Semi-Markov Models for Telemetric Monitoring of Physical Activity, Bayesian Analysis, 2023, 18 (2) : 547-577.
Sacchi, N., Ciceri, F., Bonifazi, F., Algeri, M., Gallina, A., Pollichieni, S., Raggio, E., Hadj-Amar, B., Lombardini, L., Pupella, S., Liumbruno, G., and Cardillo, M. Availability of HLA-allele-matched unrelated donors and registry size: estimation from haplotype frequency in the Italian population, Journal of Human Immunology, 2021, 82 (10) : 758-766.
Hadj-Amar, B., Finkenstädt, B., Fiecas, M. & Huckstepp, R., Identifying the Recurrence of Sleep Apnea Using a Harmonic Hidden Markov Model, Annals of Applied Statistics, 2021, 15 (3) : 1171-1993.
Hadj-Amar, B., Bayesian Analysis of Nonstationary Periodic Time Series, Ph.D. Thesis, 2020.
Hadj-Amar, B., Finkenstädt, B., Fiecas, M., Levi, F. & Huckstepp, R., Bayesian Model Search for Nonstationary Periodic Time Series, Journal of the American Statistical Association, T&M, 2020, 115 (531) : 1320–1335.

* : joint first author

Software:

AutoNOM: Automatic Nonstationary Oscillatory Modelling, a Julia software to model nonstationary periodic time series.
HHMM: Harmonic Hidden Markov Model, a Julia software to model time-varying periodic and oscillatory processes by means of a Bayesian non-parametric HHMM.
BayesianApproxHSMM: a stan software (and R utilities) to model time series and sequential data using a Bayesian approximation to Hidden semi-Markov models.
mvHMM: contains a suite of software that utilizes multivariate Bayesian hidden Markov models (HMM) for the analysis of signals from monoamine neurotransmitters.
sparseVARHSMM a stan software (and R utilities) to model temporal and contemporaneous (e.g. spatial) dependencies in multivariate time series data using a flexible VAR HSMM.
BayesRPAGP: Bayesian Random Phase-Amplitude Gaussian Process. An R software for Bayesian inference of trial-level analysis of ERP data.
CALC: Bayesian covariate-dependent anti-logistic circadian model for activity data . A stan software (and R utilities) for cohort-level Bayesian modeling of circadian rest-activity rhythms designed to integrate wearable actigraphy with demographic and clinical predictors.
sggmDAR: a Julia implementation for Bayesian estimation of sparse, time-varying Gaussian graphical models under hidden regime switching via discrete autoregressive (DAR) dynamics, featuring automatic order and complexity learning.

Page updated

Google Sites

Report abuse