Invited Sessions

In many real-world applications, complex systems play a key role. Complexity and Network Analysis share an even more common language and intrinsic features, as systems containing collections of many interacting “objects”.
The focus of the session is to capture the inherent complexity in recent application fields by using network analysis methods and models. Complex networks go beyond the simple directed-undirected and weighted-unweighted data structures described in one-mode or two-mode networks. Multilayer, multilevel and multimode networks are an example of such kinds of complex structures, allowing to represent a multitude of scenarios including different types of relationships, actors, and dynamic snapshots. The network data can themselves be enriched with additional features, such as attribute variables and meta-data, with the aim of describing real phenomena in more detail. Within this scenario, advanced clustering methods and community detection algorithms should be proposed to capture the complexity of network data structures in several application fields and in everyday life.

Model based clustering relies on the specification of finite mixture models to modell heterogeneity and detecting clusters in data.Starting from finite mixtures of univariate Normal distributions the field has rapidly evolved. Currently interest is in complex models, like mixtures of nonlinear regression models, mixtures of longitudinal and functional data or mixed continous and discrete data, and modelling issues like selection of the number of components or estimation strategies that avoid overfitting.

Increasing economic and social inequalities have recently become a challenge. Building a modern policy, diagnosing and solving social problems requires the use of various statistical methods. This session aims to be a response to the demand for better explanations of changing households’ positions in recent years. The goal of this session is to explore state-of-art developments in the application of statistical methods in household economics. We would like to focus on the classification and analysis of the distribution of household consumption, saving, wealth and indebtedness. Theoretical and applied issues of well-being, financial fragility, poverty, labour participation, and the economics of marriage are also welcome.

Factor analysis is a popular method to obtain a sparse representation of the covariance matrix of multivariate observations. This is particularly relevant for analysing the correlation among high-dimensional data. Factor models have achieved large popularity in many applied areas, such as genetics, economics, and finance. The goal of this session is to review recent research developments in the area of Bayesian factor analysis. Topics include, but are not limited to, models’ identifiability, estimation of factor loadings, sparsity, selection of factor dimensionality, efficient posterior computation, theoretical developments, as well as novel applications.

Dimensionality reduction refers to a set of techniques used in statistics and machine learning to map high-dimensional data into a low-dimensional space in order to find and visualize the main relationships characterizing the data. Cluster analysis is the grouping of objects such that objects in the same cluster are more similar to each other than they are to objects in another cluster. Both techniques have wide applicability and are ofter used in combination to project the data into a lower dimensional space to visualize the data and find patterns. This session thus aims at defining the state-of-art of their literature.

Multivariate data take a three-way form when we observe a set of variables on the same units in different occasions. Such data are also called matrix-variate because we observe a variables x occasions data matrix for each unit. Models to analyze three-way data are quite sophisticated because of the complex structure of the data. However, since three-way data emerge in different fields, such as psychology, economics, sociology, biology, chemistry etc., practitioners often request such kinds of models which can exploit the richness of the information contained in the data. The aim of this session is to present some of the recent advances in the area of three-way analysis with particular reference to models for the unsupervised classification of the units.

Recently we can easily get a large amount of data in various application fields. On the other hand, such data usually contains many outliers and would be very noisy. Moreover, the obtained data commonly has a more complex structure than classical multivariate data. Thus, new developments of robust estimation, complex data analysis (such as functional data analysis), and computational reduction are required in many fields. In addition, we reaffirm the importance of exploratory data analysis. Therefore, this session will introduce new developments in complex data analysis and computational reduction techniques.

Detection of outliers in different types of data

TBA

In a digitalised world, hectic and complex in many different facets, textual data became a fundamental source of knowledge. In several situations, people, firms and public institutions produce and exchange – on a daily basis – opinions, reports, enquiries in the form of written communications. Differently from classic numeric and categorial datasets, these raw texts need a careful pre-process before being quantitatively analysed, because the underlying information is difficult to be retrieved since natural language is inherently unstructured from a data analysis point of view. For this reason, it is necessary to implement a multi-stage process able to distil from a text collection a set of structured data that can be analysed with statistical methods. Text Mining encompasses different tasks that can satisfy many informative needs, from text summarisation to information extraction and organisation. In this framework, during the last years, scholars have adapted well known multivariate methods to the analysis of textual data or developed innovative approaches specifically focused on textual data. This specialised session aims at discussing the most recent developments in this research domain, offering to the CLADAG audience an overview about the advances concerning the algebraic models for text representation, the approaches for reducing dataset dimensionality, the categorisation of texts, both considering an unsupervised as well as a supervised standpoint.

With the advance of modern technology, more and more data are being recorded over a continuous domain resulting in functional data. Although the underlying functions are often continuous and smooth, the observed data are discrete and noisy measurements. Typical goals of FDA are function estimation (smoothing), multiple comparison of curves, functional clustering, dependence structure learning or functional regression, where the response and/or the covariates are functional. The goal of this section is to review novel contributions in FDA.

Big Data Analytics refers to all the techniques that make it possible to collect and manage Big Data and transform it into ‘Small Data’: understandable, valuable information for improving decision-making. Latent variable models are suitable statistical tools to reach this goal. These models include variables not directly observable (latent), and they are mainly used to formulate complex dependencies among observable variables, account for unobserved heterogeneity, and cluster units in separate groups. Among this class, Hidden Markov Models provide solid statistical inference procedures and are one of the most fundamental and largely applied statistical models in many different areas. This section will present recent advancements in computational and methodological issues on these topics.

TBA

Inference for fintie populations is essentially a prediction problem. In this sense, Machine learning methods are proving particularly useful in exploiting information coming from low-cost auxiliary data sources such as Administrative registers and non-probability surveys. The session aims at exploring theoretical and applied issues. The list of tentative speakers involve researchers from the Academia and National Statistical Institutes.

TBA

TBA

Latent variable models are particularly suitable to account for the dependencies between observations. The complexity of the data often requires the development of new modelling approaches that involve a large number of parameters, which are difficult to estimate using traditional methods. This session aims at presenting recent developments in model formulation and estimation, as for example composite likelihoods and regularization.

Evaluation and classification of individuals in sports have a central role in Sports Analytics literature. Team managers and coaches are learning that the statistical analysis of sports data can be crucial for effective evidence-based decision-making. Model-based methods, machine learning approaches and classical index-based studies can be used to analyse sports data. Moreover, the growing availability of disaggregated datasets (also including big data frameworks such as those collected by sensors on F1 race cars) represents a stimulus for developing new, scalable, and flexible methods. This section aims at collecting recent contributions to this field of study.

Preference judgements are predominant in data analysis, preference rankings are predominant in preference judgements. Typically, rank data consist of a set of individuals, or judges, who have ordered a set of items —or objects— according to their overall preference or some pre-specified criterion. This session covers several major aspects of the analysis of rank data: a) the association between the notions of similarity and preference by using the framework of the theory of binary relations considered as subsets of the cartesian product of the set of objects by itself; b) the aggregation of preference approvals to obtain a consensus preference-approval defined through an ad-hoc distance-based approach; c) the computational approaches utilized to find the consensus ranking when the number of items is (very) large.

Dan Ariely provocatively said that Big data is like teenage sex 🙂 In this session we would like to bring together real big data applications for prediction problems. The works of the tentative speakers cover the use of big data (mostly coming from web-scraping and sacnner) for estimation of economic indicators and businesses parameters of interest.

TBA

TBA

TBA

This session is organized by the Italian Region of the International Biometric Society with the intent to encourage the discussion on the use of statistical methods for the analysis of complex structure of data related to health problems. The development of advanced methodologies is indeed required to face the increasing proliferation of data and the new challenges which continuously rise in the clinical, omics and environmental domains. The session is composed of three invited speakers and a discussant specialized on the statistical analysis of health data.

Supervised, unsupervised and semisupervised learning methods are increasingly used in empirical investigations of financial and business problems. But there is a problem of choosing an approach, method, procedure or algorithm, which would be the most effective from an analytical and prognostic point of view and be useful for practice. The aim of the session is to give the platform for discussion and dissemination results of fruitful applications of statistical learning methods for deepening microeconomic, financial and business analyses, especially in these economically and politically unstable times.