Grant: Discovery Grants Program - Individual
University of Waterloo
$27,000.00
Approximate Inference for General Smooth Models
FY 2024-25
Summary
Statistical Science is concerned with learning about the world using imperfect information and quantifying the uncertainty in what we learn.
A wide breadth of modern science, including climate science, infectious disease epidemiology, astrophysics, microbiology and genetics rely on inferences drawn from complex statistical models. Drawing inferences from complex statistical models requires computing quantities that are impossible to compute exactly. Approximate inferences must be made.
A substantial question in the statistical sciences over the last 30 years has been how to make efficient approximate inferences based on complex statistical models, and to determine their statistical properties.
The long-term goal the proposed research program is to investigate novel approaches for making approximate inferences in complex statistical models. Students trained under the proposed research program will develop skills in advanced statistical modelling, analysis of complex data, and scientific computation, that will prepare them to succeed in data science roles in academia, industry, and government. This research will therefore contribute to the building of capacity of trained data scientists in Canada, contributing to fulfillment of Statistics Canada's Data Science Mission.
Immediate effort will be placed on making approximate inferences in a particular a broad class of models called General Smooth Models (GSMs). These models accommodate data which: are correlated in groups, time, or space; represent continuous measurements, discrete categories, counts, extremal values, point processes (such as the spread of a disease in time and space), or event times; and which may be only partially observed, or observed with error. Making useful approximate inferences based on such sophisticated models requires a suitably sophisticated approach to computation. Two prominent approaches are the Laplace Approximation (LA) which is inexpensive to compute but may be inaccurate and hence give misleading inferences, and Adaptive Quadrature (AQ) which is accurate but usually infeasibly expensive to compute.
Over the next five years, methods will be developed that improve on the accuracy of LA while retaining its efficiency; improve on the efficiency of AQ while retaining its accuracy; and extend all these methods to work with massive sets of data. These new methods will be applied to small-area mapping of HIV risk in low-resource settings for developing geographically-differentiated intervention strategies, and modelling the gene expression mechanism of bacterial dormancy for prevention of life-threatening infection.
Program
Discovery Grants Program - Individual
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-03331
Name nom
Stringer, Alexander
Department département
Statistics and Actuarial Sciences
Organization id
33
Institution établissement
University of Waterloo
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
27000
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
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
3009
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Parametric inference
Research subject fr
Inférence paramétrique
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
Approximate Inference for General Smooth Models
Keywords
computational statistics; approximate inference; convergence theory; spatial statistics; epidemiology
Application summary
Statistical Science is concerned with learning about the world using imperfect information and quantifying the uncertainty in what we learn.
A wide breadth of modern science, including climate science, infectious disease epidemiology, astrophysics, microbiology and genetics rely on inferences drawn from complex statistical models. Drawing inferences from complex statistical models requires computing quantities that are impossible to compute exactly. Approximate inferences must be made.
A substantial question in the statistical sciences over the last 30 years has been how to make efficient approximate inferences based on complex statistical models, and to determine their statistical properties.
The long-term goal the proposed research program is to investigate novel approaches for making approximate inferences in complex statistical models. Students trained under the proposed research program will develop skills in advanced statistical modelling, analysis of complex data, and scientific computation, that will prepare them to succeed in data science roles in academia, industry, and government. This research will therefore contribute to the building of capacity of trained data scientists in Canada, contributing to fulfillment of Statistics Canada's Data Science Mission.
Immediate effort will be placed on making approximate inferences in a particular a broad class of models called General Smooth Models (GSMs). These models accommodate data which: are correlated in groups, time, or space; represent continuous measurements, discrete categories, counts, extremal values, point processes (such as the spread of a disease in time and space), or event times; and which may be only partially observed, or observed with error. Making useful approximate inferences based on such sophisticated models requires a suitably sophisticated approach to computation. Two prominent approaches are the Laplace Approximation (LA) which is inexpensive to compute but may be inaccurate and hence give misleading inferences, and Adaptive Quadrature (AQ) which is accurate but usually infeasibly expensive to compute.
Over the next five years, methods will be developed that improve on the accuracy of LA while retaining its efficiency; improve on the efficiency of AQ while retaining its accuracy; and extend all these methods to work with massive sets of data. These new methods will be applied to small-area mapping of HIV risk in low-resource settings for developing geographically-differentiated intervention strategies, and modelling the gene expression mechanism of bacterial dormancy for prevention of life-threatening infection.