Grant: Discovery Launch Supplement
McGill University
$12,500.00
Statistical learning for high-dimensional survey sampling
FY 2024-25
Summary
Complex surveys are essential in providing information for policymakers, the general public, and many scientists in other areas, such as public health and social science research. Survey statisticians now have vast information about individuals, businesses, etc. This abundance of data enables the application of sophisticated machine learning algorithms for different purposes, including the treatment of both unit and item nonresponse, model-assisted estimation, and small area estimation. The use of complex machine learning methods in surveys is a relatively unexplored research domain. Also, the existing literature on predictive models has mainly relied on the assumption that the number of predictors is small with respect to the sample size, an assumption not always satisfied in practice.
The research program presented in this proposal aims to develop efficient and theoretically sound estimation strategies based on statistical learning procedures for high-dimensional survey data, including, at their core, the treatment of missing data. Within this program, several intertwined projects will be investigated. This includes the study of the high-dimensional behaviors of standard point and variance survey estimators, the development and study of model selection and model-aggregation procedures, as well as the development of new survey estimators based on statistical learning procedures (e.g., boosting).
This research will likely impact academia and industry in three important ways: first, analyzing and communicating the undesirable effects of high-dimensional covariates with traditional survey methods will raise practitioners' awareness. Second, suggesting new methodological developments robust to high-dimensional scenarios will give practitioners tools when traditional methods become unsafe. Third, the development of aggregation and estimator selection procedures will give practitioners new tools to choose which estimation procedure to use. Overall, this will likely result in improved efficiency for survey estimation strategies. These more efficient survey strategies will help researchers provide more reliable and representative statistics for understanding social, economic, and health trends, identifying disparities, and informing policies promoting inclusivity and equity.
Program
Discovery Launch Supplement
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
DGECR-2024-00457
Name nom
Dagdoug, Mohamed Mehdi
Department département
Department of Mathematics and Statistics
Organization id
61
Institution établissement
McGill University
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
12500
Program id
DGECR
Program name en
Discovery Launch Supplement
Program name fr
Supplément Tremplin vers la découverte
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
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
Research subject code
3003
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Survey methodology
Research subject fr
Techniques de sondage
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 learning for high-dimensional survey sampling
Keywords
survey sampling; statistical learning; high-dimensional statistics; missing data; aggregation methods; estimator selection; variance estimation
Application summary
Complex surveys are essential in providing information for policymakers, the general public, and many scientists in other areas, such as public health and social science research. Survey statisticians now have vast information about individuals, businesses, etc. This abundance of data enables the application of sophisticated machine learning algorithms for different purposes, including the treatment of both unit and item nonresponse, model-assisted estimation, and small area estimation. The use of complex machine learning methods in surveys is a relatively unexplored research domain. Also, the existing literature on predictive models has mainly relied on the assumption that the number of predictors is small with respect to the sample size, an assumption not always satisfied in practice.
The research program presented in this proposal aims to develop efficient and theoretically sound estimation strategies based on statistical learning procedures for high-dimensional survey data, including, at their core, the treatment of missing data. Within this program, several intertwined projects will be investigated. This includes the study of the high-dimensional behaviors of standard point and variance survey estimators, the development and study of model selection and model-aggregation procedures, as well as the development of new survey estimators based on statistical learning procedures (e.g., boosting).
This research will likely impact academia and industry in three important ways: first, analyzing and communicating the undesirable effects of high-dimensional covariates with traditional survey methods will raise practitioners' awareness. Second, suggesting new methodological developments robust to high-dimensional scenarios will give practitioners tools when traditional methods become unsafe. Third, the development of aggregation and estimator selection procedures will give practitioners new tools to choose which estimation procedure to use. Overall, this will likely result in improved efficiency for survey estimation strategies. These more efficient survey strategies will help researchers provide more reliable and representative statistics for understanding social, economic, and health trends, identifying disparities, and informing policies promoting inclusivity and equity.