Sparse Composite Likelihood Inference in Count Time Series

About SCouT

In public health surveillance, accurate monitoring, analysis and prediction of disease counts are fundamental for effective and timely detection of disease outbreaks. Surveillance data usually appear in form of count time series.
Statistical analysis of this type of data is challenging because it combines difficulties related to modelling discrete data with the need to account for serial dependence.

The EU funded project SCouT aims at developing statistical tools able to balance competing needs for timeliness, model simplicity and effectiveness in outbreak detection.
The basic statistical tool used in SCouT is composite likelihood that follow the `divide and conquer principle’: split a complicated problem into simpler subproblems and then aggregate.



  • Pedeli, X. and Karlis, D. (2018). An integer-valued time series model for multivariate surveillance, arXiv:1805.08561
  • Pedeli, X. and Varin, C. (2018). Pairwise likelihood estimation of latent autoregressive count models, arXiv:1805.10865

Public statistical package


Indirect pairwise fitting of latent autoregressive and moving average models

  • Conference: International Conference of the ERCIM WG on Computational and Methodological Statistics (CMStatistics 2017), 16-18 December, 2017
  • Authors: Guido Masarotto, Xanthi Pedeli and Cristiano Varin

The Pairwise Expectation Maximization Algorithm for Fitting Parameter-Driven Models

  • Conference: International Workshop on Statistical Modelling (IWSM 2017), 2-7 July 2017, Groningen, Netherlands
  • Authors: Xanthi Pedeli and Cristiano Varin
file pdfSlides363 K

Pairwise likelihood inference for parameter-driven models

  • Conference: SIS 2017 Conference, 28-30 June 2017, Florence, Italy
  • Authors: Xanthi Pedeli and Cristiano Varin
file pdfSlides759 K

Aspects of composite likelihood inference in time series models

  • Conference: International Conference of the ERCIM WG on Computational and Methodological Statistics (CMStatistics 2016), 9-11 December 2016, Seville, Spain
  • Authors: Xanthi Pedeli and Cristiano Varin
file pdfSlides163 K


Xanthi Xanthipi Pedeli
Marie Skłodowska-Curie fellow

Xanthi Pedeli is a Marie Skłodowska-Curie research fellow at Ca’ Foscari University of Venice.
Her research interests include statistical modelling and inference for time series, count data, multivariate models and biostatistics.
Citations profile on Google Scholar.

Cristiano Varin

Cristiano Varin is Associate Professor of Statistics at Ca' Foscari University of Venice.
His research is focused on composite likelihood inference, meta-analysis and paired-comparison modelling. A special interest is the development of open source statistical software in the R programming environment.
Curriculum vitae and citations profile on Google Scholar.

Roland Fried
Secondment supervisor

Roland Fried is Full Professor of Statistics in Biosciences at the Technical University of Dortmund.
His areas of expertise include biostatistics, modelling of spatial data and time series, online monitoring, robust signal extraction and change point detection.
Curriculum vitae ad citations profile on Google Scholar.

Dimitris Karlis
Scientific Collaborator

Professor of Statistics, Athens University of Economics and Business

Aristidis Nikoloulopoulos
Scientific Collaborator

Senior Lecturer in Statistics, University of East Anglia

Guido Masarotto
Scientific Collaborator

Full professor of Statistics, University of Padua