Grant: Discovery Grants Program - Individual
University of Toronto
$40,000.00
Statistical modelling with positive dependence
EF 2024-25
Résumé
In many complex systems the variables are strongly positively related. This may happen, for example, when there is common factor that drives the behaviour of all the observables, or, when the observables evolved from a common source. Typical examples of such processes include stock market data, survey data in psychology, medical data related to a particular disease syndrome, multineuronal activity patterns in brain data, or evolutionary biology. In these applications it is natural then to incorporate positive dependence explicitly into the statistical model.
Positive dependence has been classically studied in probability theory, statistical physics, and statistics. However, it has been only recently observed that it may provide powerful tools for modern multivariate statistical modelling leading in some applications to superior statistical performance and sparsity. It is now important to understand better the complex relation between positive dependence and statistical modelling.
The focus of this research program is on the theoretical and computational aspects of statistical modelling with positive dependence. Positive dependence is understood very broadly and it can be relatively weak, assuming only positive correlations in the system, or relatively strong, assuming nonnegative partial correlations. I aim at advancing the recently developed theory in two main directions. First, it is important to provide reliable tools that can be used within the statistical practice in order to test and use positive dependence in a more automatized way. Second, my goal is to study positive dependence in the context of causal analysis and directed graphical models, which leads to considering directional notions of positive dependence. Causal analysis with directed graphs provides many examples of systems that are clearly positive dependent.
This proposal builds upon a recent line of work that mostly focused on Gaussian distributions with nonnegative partial correlations and related ideas. Although the developed methods perform extremely well in selected applications, the first objective of this proposed research program is critical for developing meaningful and widely applicable theory and for transforming statistical practice. Advancing our understanding of modelling with positive dependence may lead to new, more robust methods to analyze data in specific contexts. The second objective of the proposed program has also a strong motivation. Indeed, inferring causal relationships has always been one of the main objectives of science. Examples in today's world include, but are not limited to, inferring potential causes of cancer, the effect of gene manipulation, and much more. As in the case of earlier work, the study of positivity on directed graphs is likely to lead to advances in statistical theory, which is of general interest for the mathematical statistics community.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2023-03481
Name nom
Zwiernik, Piotr
Department département
Statistical Sciences
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2023
Award amount
40000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1200
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Advancement of knowledge
Area of application fr
Avancement des connaissances
Research subject code
3011
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Multivariate analysis
Research subject fr
Analyse multivariée
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Statistical modelling with positive dependence
Keywords
positive dependence; graphical models; bayesian networks; positive regression; positive pca; m-matrices; structure learning; convex optimization; covariance matrix estimation; association
Application summary
In many complex systems the variables are strongly positively related. This may happen, for example, when there is common factor that drives the behaviour of all the observables, or, when the observables evolved from a common source. Typical examples of such processes include stock market data, survey data in psychology, medical data related to a particular disease syndrome, multineuronal activity patterns in brain data, or evolutionary biology. In these applications it is natural then to incorporate positive dependence explicitly into the statistical model.
Positive dependence has been classically studied in probability theory, statistical physics, and statistics. However, it has been only recently observed that it may provide powerful tools for modern multivariate statistical modelling leading in some applications to superior statistical performance and sparsity. It is now important to understand better the complex relation between positive dependence and statistical modelling.
The focus of this research program is on the theoretical and computational aspects of statistical modelling with positive dependence. Positive dependence is understood very broadly and it can be relatively weak, assuming only positive correlations in the system, or relatively strong, assuming nonnegative partial correlations. I aim at advancing the recently developed theory in two main directions. First, it is important to provide reliable tools that can be used within the statistical practice in order to test and use positive dependence in a more automatized way. Second, my goal is to study positive dependence in the context of causal analysis and directed graphical models, which leads to considering directional notions of positive dependence. Causal analysis with directed graphs provides many examples of systems that are clearly positive dependent.
This proposal builds upon a recent line of work that mostly focused on Gaussian distributions with nonnegative partial correlations and related ideas. Although the developed methods perform extremely well in selected applications, the first objective of this proposed research program is critical for developing meaningful and widely applicable theory and for transforming statistical practice. Advancing our understanding of modelling with positive dependence may lead to new, more robust methods to analyze data in specific contexts. The second objective of the proposed program has also a strong motivation. Indeed, inferring causal relationships has always been one of the main objectives of science. Examples in today's world include, but are not limited to, inferring potential causes of cancer, the effect of gene manipulation, and much more. As in the case of earlier work, the study of positivity on directed graphs is likely to lead to advances in statistical theory, which is of general interest for the mathematical statistics community.