Grant: Discovery Grants Program - Individual
McGill University
$67,000.00
Generalizing Bayes: approaches to prior-to-posterior updating within and beyond the Bayesian paradigm
FY 2024-25
Summary
Much of modern society is data-driven: developments in data collection methods mean that huge amounts of information can be used to inform optimal decisions, such as the development of treatment regimen or policies in healthcare, or prediction of future outcomes on the basis of current data. From a statistical viewpoint, it has long been known that the best decisions are made using the Bayesian paradigm, a technique that updates prior knowledge and current data to represent posterior uncertainty in a coherent fashion. My long-term research goal is to develop statistical procedures and computational tools to better facilitate the use of data in decision-making. The research is theoretical and methodological in the mathematical sciences, and does not involve Indigenous communities, societies, or individuals, and does not qualify as Indigenous research.
The modern implementation of Bayesian statistical methods has relied upon the long-standing mathematical result known as Bayes theorem which gives a formula indicating how prior knowledge should be updated in light of new data collected in a study. Over the last 30 years, computational advances have revolutionized how widely Bayesian methods can be used to handle large data sets. However, traditional Bayesian methods are tied to specific modelling assumptions that describe how the observed data are generated, and this can be quite limiting. My proposed research aims to use the Bayesian method, but to generalize the types of model specification that can be used, and to relax the strictures of the traditional Bayesian approach.
In my last Discovery Grant, I developed theory and methods to perform statistical analyses based on minimal modelling assumptions. This contributed to the statistical literature, and had impact in practical applications, and also the training of highly qualified personnel. In the next phase of this program of research, my aim is to develop more flexible approaches to prior-to-posterior information updating that are still probabilistic and rely on the observed data in a coherent manner but are more flexible. I will use developments in neural networks, which are large and complex models that have been designed for predicting outcomes in arbitrarily complex problems; if the challenge is inference (that is, understanding the structure of a data generating process), then neural networks are much more limited. My research aims to use the prediction capabilities of these models to carry out Bayesian updating in a fully inferential framework.
My research will contribute to the statistical and computational literature, but also produce new tools for researchers so that they can use Bayesian methodology in their own fields, thereby contributing to scientific research at large. I will also involve students and other trainees in this work, supervising undergraduate and graduate students who will go on to become independent researchers or join the workforce in another capacity.
Program
Discovery Grants Program - Individual
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2024-04802
Name nom
Stephens, David
Department département
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
67000
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
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
Generalizing Bayes: approaches to prior-to-posterior updating within and beyond the Bayesian paradigm
Keywords
bayesian inference; updating rules; causal inference; computational methods; prediction
Application summary
Much of modern society is data-driven: developments in data collection methods mean that huge amounts of information can be used to inform optimal decisions, such as the development of treatment regimen or policies in healthcare, or prediction of future outcomes on the basis of current data. From a statistical viewpoint, it has long been known that the best decisions are made using the Bayesian paradigm, a technique that updates prior knowledge and current data to represent posterior uncertainty in a coherent fashion. My long-term research goal is to develop statistical procedures and computational tools to better facilitate the use of data in decision-making. The research is theoretical and methodological in the mathematical sciences, and does not involve Indigenous communities, societies, or individuals, and does not qualify as Indigenous research.
The modern implementation of Bayesian statistical methods has relied upon the long-standing mathematical result known as Bayes theorem which gives a formula indicating how prior knowledge should be updated in light of new data collected in a study. Over the last 30 years, computational advances have revolutionized how widely Bayesian methods can be used to handle large data sets. However, traditional Bayesian methods are tied to specific modelling assumptions that describe how the observed data are generated, and this can be quite limiting. My proposed research aims to use the Bayesian method, but to generalize the types of model specification that can be used, and to relax the strictures of the traditional Bayesian approach.
In my last Discovery Grant, I developed theory and methods to perform statistical analyses based on minimal modelling assumptions. This contributed to the statistical literature, and had impact in practical applications, and also the training of highly qualified personnel. In the next phase of this program of research, my aim is to develop more flexible approaches to prior-to-posterior information updating that are still probabilistic and rely on the observed data in a coherent manner but are more flexible. I will use developments in neural networks, which are large and complex models that have been designed for predicting outcomes in arbitrarily complex problems; if the challenge is inference (that is, understanding the structure of a data generating process), then neural networks are much more limited. My research aims to use the prediction capabilities of these models to carry out Bayesian updating in a fully inferential framework.
My research will contribute to the statistical and computational literature, but also produce new tools for researchers so that they can use Bayesian methodology in their own fields, thereby contributing to scientific research at large. I will also involve students and other trainees in this work, supervising undergraduate and graduate students who will go on to become independent researchers or join the workforce in another capacity.