Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

Carleton University

$31,000.00

Efficient statistical inference in sparse models

EF 2024-25

Résumé
The last three decades have seen the development of statistical models that are infinite (or 'high') dimensional. The main target of statistical inference in high-dimensional models is a function or an infinite vector. In view of modern computational techniques, such models are tractable and attractive in statistical practice. The goal of this research project is to elaborate statistically efficient and computationally feasible methods that would allow scientists and practitioners to make statistical inference in sparse high-dimensional models and apply the proposed methods for analyzing real-life data. It is often difficult to treat high-dimensional problems and provide practical procedures with nearly exact theoretical properties. Nevertheless, applications of some advanced statistical theories (if possible) may lead to elegant and efficient solutions to various high-dimensional problems under sparsity assumptions. In the problems of detecting and recovering sparse high-dimensional signals, both asymptotic and non-asymptotic analyses are to be carried out for some challenging (non-additive) regression schemes. The main strategy to be employed is adaptive thresholding. Various techniques for adaptation of a statistical procedure to unknown model parameters, including Bonferroni's method and Lepski's method, will be used. It is often the case that solving sparse recovery problems in nonparametric settings requires the use of relevant hypothesis testing results. These results are available in the literature only in part and, in case of new regression schemes, are to be derived. Depending on the level of signal strength, sparse recovery may or may not be possible. Conditions for the possibility and impossibility of sparse signal recovery are to be established under different scenarios. These conditions may be used by a practitioner who needs to design experiments as benchmarks. The problems of detecting and identifying sparse signal segments in high-dimensional compositional data, with possible applications to microbiome differential abundance analysis will also be addressed. Quantitative comparison of microbial composition from different populations is a fundamental problem in many microbiome studies. The questions of interest for us are how to effectively detect the sparse composition difference between the populations and to identify the specific location of a small fraction of the informative (differentially abundant) bacterial taxa in a microbial community when the null hypothesis of no difference is rejected. These problems present many challenges for both theoretical and applied scientists, and therefore should be tackled jointly by researchers and practitioners. The project aims at contributing to the mathematical theory of statistical inference in high-dimensional models. Analytic findings of the project are to be supported by simulation studies and to be applied to synthetic data and to microbiome abundance data.
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-2024-05354
Name nom
Stepanova, Natalia
Department département
School of Mathematics and Statistics
Organization id
23
Institution établissement
Carleton University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
31000
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
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
3000
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Statistics and probability
Research subject fr
Statistique et probabilité
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
Efficient statistical inference in sparse models
Keywords
high-dimensional regression; dimensionality reduction; nonparametric hypothesis testing; nonparametric estimation; minimaxity and adaptivity; variable selection in sparse models; infererence for sparse compositional data; gaussian white noise; sparse signal detection; sparse signal recovery
Application summary
The last three decades have seen the development of statistical models that are infinite (or 'high') dimensional. The main target of statistical inference in high-dimensional models is a function or an infinite vector. In view of modern computational techniques, such models are tractable and attractive in statistical practice. The goal of this research project is to elaborate statistically efficient and computationally feasible methods that would allow scientists and practitioners to make statistical inference in sparse high-dimensional models and apply the proposed methods for analyzing real-life data. It is often difficult to treat high-dimensional problems and provide practical procedures with nearly exact theoretical properties. Nevertheless, applications of some advanced statistical theories (if possible) may lead to elegant and efficient solutions to various high-dimensional problems under sparsity assumptions. In the problems of detecting and recovering sparse high-dimensional signals, both asymptotic and non-asymptotic analyses are to be carried out for some challenging (non-additive) regression schemes. The main strategy to be employed is adaptive thresholding. Various techniques for adaptation of a statistical procedure to unknown model parameters, including Bonferroni's method and Lepski's method, will be used. It is often the case that solving sparse recovery problems in nonparametric settings requires the use of relevant hypothesis testing results. These results are available in the literature only in part and, in case of new regression schemes, are to be derived. Depending on the level of signal strength, sparse recovery may or may not be possible. Conditions for the possibility and impossibility of sparse signal recovery are to be established under different scenarios. These conditions may be used by a practitioner who needs to design experiments as benchmarks. The problems of detecting and identifying sparse signal segments in high-dimensional compositional data, with possible applications to microbiome differential abundance analysis will also be addressed. Quantitative comparison of microbial composition from different populations is a fundamental problem in many microbiome studies. The questions of interest for us are how to effectively detect the sparse composition difference between the populations and to identify the specific location of a small fraction of the informative (differentially abundant) bacterial taxa in a microbial community when the null hypothesis of no difference is rejected. These problems present many challenges for both theoretical and applied scientists, and therefore should be tackled jointly by researchers and practitioners. The project aims at contributing to the mathematical theory of statistical inference in high-dimensional models. Analytic findings of the project are to be supported by simulation studies and to be applied to synthetic data and to microbiome abundance data.