Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

York University

$28,000.00

Graphical Gaussian and hierarchical discrete models

FY 2014-15

Summary
This research proposal is in two parts. The first part concerns graphical Gaussian models, the search for conjugate priors for the covariance parameter of various models, and their properties, in particular (a) the study of duality between conjugate priors for covariance graph and precision graph models(b) the behaviour of the eigenvalues of the G-Wishart prior in the decomposable case. These properties are studied with a view to improving covariance estimation, especially in the case high-dimensional data and relatively few data points. Covariance estimation under these conditions is one of the challenges of modern multivariate statistics.The second part concerns discrete models, in particular,(a) the geometry of the hierarchical loglinear model through its marginal polytope and its facets, (b) the development of conjugate priors for discrete models other than the loglinear models (e.g. marginal independence models),(c) the characterization of the hyper Dirichlet,(d) topics in missing data problems in a Bayesian framework.Within the topic of the geometry of the loglinear model, we are particularly interested in using some of our previous results to give accurate approximations to the distribution of certain model selection statistics and an approximation of a general model by a ""best"" simpler decomposable one. We also want to investigate the role of the characteristic function of the marginal polytope in achieving some of these approximations.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
8947-2011
Name nom
Massam, Hélène
Department département
Mathematics and Statistics
Organization id
38
Institution établissement
York University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2014
Competition year année de concours
2011
Award amount
28000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1211
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Medical and health sciences
Area of application fr
Sciences médicales et de la santé
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
Partie
4
Nb partie
5
Application title
Graphical Gaussian and hierarchical discrete models
Keyword
NSERC
Application summary
This research proposal is in two parts. The first part concerns graphical Gaussian models, the search for conjugate priors for the covariance parameter of various models, and their properties, in particular (a) the study of duality between conjugate priors for covariance graph and precision graph models(b) the behaviour of the eigenvalues of the G-Wishart prior in the decomposable case. These properties are studied with a view to improving covariance estimation, especially in the case high-dimensional data and relatively few data points. Covariance estimation under these conditions is one of the challenges of modern multivariate statistics.The second part concerns discrete models, in particular,(a) the geometry of the hierarchical loglinear model through its marginal polytope and its facets, (b) the development of conjugate priors for discrete models other than the loglinear models (e.g. marginal independence models),(c) the characterization of the hyper Dirichlet,(d) topics in missing data problems in a Bayesian framework.Within the topic of the geometry of the loglinear model, we are particularly interested in using some of our previous results to give accurate approximations to the distribution of certain model selection statistics and an approximation of a general model by a ""best"" simpler decomposable one. We also want to investigate the role of the characteristic function of the marginal polytope in achieving some of these approximations.