Grant: Discovery Grants Program - Individual
The University of Western Ontario
$33,000.00
Statistical learning of health data with complex features
EF 2023-24
Résumé
The advancement of modern technology brings in new techniques to collect rich sources of data, which, in turn, stimulates unprecedented scientific investigations from angles different from traditional perspectives. The complex features introduced by the new techniques present new challenges to traditional statistical learning methods. Addressing the emerging issues becomes compulsory and urgent in developing innovative statistical learning methods. Building effective prediction models is of primary interest in medical and health studies. Many existing methods do not hold for complex featured observations, such as those with text medical notes, noisy data, and the lack of expert resources in providing accurate patient outcomes. To make effective use of available resources of data, it is imperative to explore novel approaches to model and analyze data with complex features. This research aims to develop novel statistical learning methods to deal with data with complex features. Specifically, I plan to explore: 1) dynamic treatment regimes with noisy data; and 2) active learning for text medical data and effective prediction of patient diagnosis outcome. In Stream one, I plan to develop valid methods to address the effects of measurement error in both covariates and outcomes. The impact of mismeasured variables on the identification of optimal treatment and the methodology of adjusting the effect of measurement error in dynamic treatment regimes will be investigated. Inference and variable selection will be explored when measurement error is unavoidable in covariates. Model-based and outcome-weighted value functions will be investigated in dynamic treatment regimes, and ensemble strategies such as boosting will be employed to develop efficient computational algorithms. When the medical diagnosis is difficult to obtain, or the medical resources of transferring medical text notes into numerical values are involved in predictive models, effective use of the resources is an important issue. In Stream two of my research, I plan to develop methodologies in active learning to help text note transferring and patient outcome diagnosis to tackle the problem of limited resources. By selecting a small portion of typical cases using the active learning strategies, I will develop viable prediction models to achieve a similar prediction accuracy as if all the patients are diagnosed. The proposed methodology development will lead to valuable new insights into making use of medical data with complex features. The research is expected to offer novel additions to the fields of statistical science and data science. The developed methods can be utilized to handle complex featured data from diverse areas, such as economics, engineering, and sociology.
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-2023-05539
Name nom
He, Wenqing
Department département
Statistical and Actuarial Sciences
Organization id
36
Institution établissement
The University of Western Ontario
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
33000
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
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
3004
Research subject group en
Statistics and probability
Research subject group fr
Applied statistics
Research subject en
Statistique et probabilité
Research subject fr
Statistique appliquée
Application title
Statistical learning of health data with complex features
Keyword
Active learning; Dynamic treatment regimes; Miaasing data and Measurement error; Outcome weighted learning; Statistical learning
Application summary
The advancement of modern technology brings in new techniques to collect rich sources of data, which, in turn, stimulates unprecedented scientific investigations from angles different from traditional perspectives. The complex features introduced by the new techniques present new challenges to traditional statistical learning methods. Addressing the emerging issues becomes compulsory and urgent in developing innovative statistical learning methods. Building effective prediction models is of primary interest in medical and health studies. Many existing methods do not hold for complex featured observations, such as those with text medical notes, noisy data, and the lack of expert resources in providing accurate patient outcomes. To make effective use of available resources of data, it is imperative to explore novel approaches to model and analyze data with complex features. This research aims to develop novel statistical learning methods to deal with data with complex features. Specifically, I plan to explore: 1) dynamic treatment regimes with noisy data; and 2) active learning for text medical data and effective prediction of patient diagnosis outcome. In Stream one, I plan to develop valid methods to address the effects of measurement error in both covariates and outcomes. The impact of mismeasured variables on the identification of optimal treatment and the methodology of adjusting the effect of measurement error in dynamic treatment regimes will be investigated. Inference and variable selection will be explored when measurement error is unavoidable in covariates. Model-based and outcome-weighted value functions will be investigated in dynamic treatment regimes, and ensemble strategies such as boosting will be employed to develop efficient computational algorithms. When the medical diagnosis is difficult to obtain, or the medical resources of transferring medical text notes into numerical values are involved in predictive models, effective use of the resources is an important issue. In Stream two of my research, I plan to develop methodologies in active learning to help text note transferring and patient outcome diagnosis to tackle the problem of limited resources. By selecting a small portion of typical cases using the active learning strategies, I will develop viable prediction models to achieve a similar prediction accuracy as if all the patients are diagnosed. The proposed methodology development will lead to valuable new insights into making use of medical data with complex features. The research is expected to offer novel additions to the fields of statistical science and data science. The developed methods can be utilized to handle complex featured data from diverse areas, such as economics, engineering, and sociology.