Sahir Bhatnagar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sahir Bhatnagar has a total of 155 co-authors. Here are the top co-authors by number of shared publications:
- Celia Margaret T. Greenwood
Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Canada
8 shared publications - Karim Oualkacha
Département de mathématiques, Université du Québec à Montréal, Montréal, Canada
4 shared publications - Reza Forghani
McGill University
3 shared publications - Tianyuan Lu
Department of Population Health Sciences, School of Medicine and Public Health, University of Wisconsin-Madison, Madison, WI, USA
3 shared publications - Yi Yang
Department of Mathematics and Statistics, McGill University, Montréal, Québec, Canada
3 shared publications - Caroline Reinhold
Department of Radiology, McGill University Health Centre, Montreal, Canada
2 shared publications
Latest Publications
- 2 January 2024
- Published by Springer Nature in Statistics and Computing
- Vol. 34 (1) , 1-20
Accelerated gradient methods for sparse statistical learning with nonconvex penalties
- 2 January 2024
- Published by Springer Nature in Statistics and Computing
- Vol. 34 (1) , 1-20
Accelerated gradient methods for sparse statistical learning with nonconvex penalties
- Open Access
Variable Selection for Individualized Treatment Rules with Discrete Outcomes
An individualized treatment rule (ITR) is a decision rule that aims to improve individual patients health outcomes by recommending optimal treatments according to patients specific information. In observational studies, collected data may contain many variables that are irrelevant for making treatment decisions. Including all available variables in the statistical model for the ITR could yield a loss of efficiency and an unnecessarily complicated treatment rule, which is difficult for physicians to interpret or implement. Thus, a data-driven approach to select important tailoring variables with the aim of improving the estimated decision rules is crucial. While there is a growing body of literature on selecting variables in ITRs with continuous outcomes, relatively few methods exist for discrete outcomes, which pose additional computational challenges even in the absence of variable selection. In this paper, we propose a variable selection method for ITRs with discrete outcomes. We show theoretically and empirically that our approach has the double robustness property, and that it compares favorably with other competing approaches. We illustrate the proposed method on data from a study of an adaptive web-based stress management tool to identify which variables are relevant for tailoring treatment.
Open AccessVariable Selection for Individualized Treatment Rules with Discrete Outcomes
An individualized treatment rule (ITR) is a decision rule that aims to improve individual patients health outcomes by recommending optimal treatments according to patients specific information. In observational studies, collected data may contain many variables that are irrelevant for making treatment decisions. Including all available variables in the statistical model for the ITR could yield a loss of efficiency and an unnecessarily complicated treatment rule, which is difficult for physicians to interpret or implement. Thus, a data-driven approach to select important tailoring variables with the aim of improving the estimated decision rules is crucial. While there is a growing body of literature on selecting variables in ITRs with continuous outcomes, relatively few methods exist for discrete outcomes, which pose additional computational challenges even in the absence of variable selection. In this paper, we propose a variable selection method for ITRs with discrete outcomes. We show theoretically and empirically that our approach has the double robustness property, and that it compares favorably with other competing approaches. We illustrate the proposed method on data from a study of an adaptive web-based stress management tool to identify which variables are relevant for tailoring treatment.
- Open Access
- 27 January 2023
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 39 (2)
Efficient penalized generalized linear mixed models for variable selection and genetic risk prediction in high-dimensional data
Sparse regularized regression methods are now widely used in genome-wide association studies (GWAS) to address the multiple testing burden that limits discovery of potentially important predictors. Linear mixed models (LMMs) have become an attractive alternative to principal components (PC) adjustment to account for population structure and relatedness in high-dimensional penalized models. However, their use in binary trait GWAS rely on the invalid assumption that the residual variance does not depend on the estimated regression coefficients. Moreover, LMMs use a single spectral decomposition of the covariance matrix of the responses, which is no longer possible in generalized linear mixed models (GLMMs). We introduce a new method called pglmm, a penalized GLMM that allows to simultaneously select genetic markers and estimate their effects, accounting for between-individual correlations and binary nature of the trait. We develop a computationally efficient algorithm based on penalized quasi-likelihood (PQL) estimation that allows to scale regularized mixed models on high-dimensional binary trait GWAS. We show through simulations that when the dimensionality of the relatedness matrix is high, penalized LMM and logistic regression with PC adjustment fail to select important predictors, and have inferior prediction accuracy compared to pglmm. Further, we demonstrate through the analysis of two polygenic binary traits in a subset of 6731 related individuals from the UK Biobank data with 320K SNPs that our method can achieve higher predictive performance, while also selecting fewer predictors than a sparse regularized logistic lasso with PC adjustment. Our Julia package PenalizedGLMM.jl is publicly available on github: https://github.com/julstpierre/PenalizedGLMM. Supplementary data are available at Bioinformatics online.
- 27 January 2023
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 39 (2)
Open AccessEfficient penalized generalized linear mixed models for variable selection and genetic risk prediction in high-dimensional data
Sparse regularized regression methods are now widely used in genome-wide association studies (GWAS) to address the multiple testing burden that limits discovery of potentially important predictors. Linear mixed models (LMMs) have become an attractive alternative to principal components (PC) adjustment to account for population structure and relatedness in high-dimensional penalized models. However, their use in binary trait GWAS rely on the invalid assumption that the residual variance does not depend on the estimated regression coefficients. Moreover, LMMs use a single spectral decomposition of the covariance matrix of the responses, which is no longer possible in generalized linear mixed models (GLMMs). We introduce a new method called pglmm, a penalized GLMM that allows to simultaneously select genetic markers and estimate their effects, accounting for between-individual correlations and binary nature of the trait. We develop a computationally efficient algorithm based on penalized quasi-likelihood (PQL) estimation that allows to scale regularized mixed models on high-dimensional binary trait GWAS. We show through simulations that when the dimensionality of the relatedness matrix is high, penalized LMM and logistic regression with PC adjustment fail to select important predictors, and have inferior prediction accuracy compared to pglmm. Further, we demonstrate through the analysis of two polygenic binary traits in a subset of 6731 related individuals from the UK Biobank data with 320K SNPs that our method can achieve higher predictive performance, while also selecting fewer predictors than a sparse regularized logistic lasso with PC adjustment. Our Julia package PenalizedGLMM.jl is publicly available on github: https://github.com/julstpierre/PenalizedGLMM. Supplementary data are available at Bioinformatics online.
- Open Access
- 26 January 2023
- Published by Public Library of Science (PLoS) in PLOS Global Public Health
- Vol. 3 (1) , e0001502
Socioeconomic factors impact the risk of HIV acquisition in the township population of South Africa: A Bayesian analysis
With a prevalence almost twice as high as the national average, people living in South African townships are particularly impacted by the HIV epidemic. Yet, it remains unclear how socioeconomic factors impact the risk of HIV infection within township populations. Our objective was to estimate the extent to which socioeconomic factors (dwelling situation, education, employment status, and monthly income) explain the risk of HIV in South African township populations, after controlling for behavioural and individual risk factors. Using Bayesian logistic regression, we analysed secondary data from a quasi-randomised trial which recruited participants (N = 3095) from townships located across three subdistricts of Cape Town. We controlled for individual factors (age, sex, marital status, testing history, HIV exposure, comorbidities, and tuberculosis infection) and behavioural factors (unprotected sex, sex with multiple partners, with sex workers, with a partner living with HIV, under the influence of alcohol or drugs), and accounted for the uncertainty due to missing data through multiple imputation. We found that residing in informal dwellings and not having post-secondary education increased the odds of HIV (aOR, 89% CrI: 1.34, 1.07–1.68 and 1.82, 1.29–2.61, respectively), after controlling for subdistrict of residence, individual, and behavioural factors. Additionally, our results suggest different pathways for how socioeconomic status (SES) affect HIV infection in males and female participants: while socioeconomic factors associated with lower SES seem to be associated with a decreased likelihood of having recently sough HIV testing among male participants, they are associated with increased sexual risk taking which, among female participants, increase the risk of HIV. Our analyses demonstrate that social determinants of health are at the root of the HIV epidemic and affect the risk of HIV in multiple ways. These findings stress the need for the deployment of programs that specifically address social determinants of health.
- 26 January 2023
- Published by Public Library of Science (PLoS) in PLOS Global Public Health
- Vol. 3 (1) , e0001502
Open AccessSocioeconomic factors impact the risk of HIV acquisition in the township population of South Africa: A Bayesian analysis
With a prevalence almost twice as high as the national average, people living in South African townships are particularly impacted by the HIV epidemic. Yet, it remains unclear how socioeconomic factors impact the risk of HIV infection within township populations. Our objective was to estimate the extent to which socioeconomic factors (dwelling situation, education, employment status, and monthly income) explain the risk of HIV in South African township populations, after controlling for behavioural and individual risk factors. Using Bayesian logistic regression, we analysed secondary data from a quasi-randomised trial which recruited participants (N = 3095) from townships located across three subdistricts of Cape Town. We controlled for individual factors (age, sex, marital status, testing history, HIV exposure, comorbidities, and tuberculosis infection) and behavioural factors (unprotected sex, sex with multiple partners, with sex workers, with a partner living with HIV, under the influence of alcohol or drugs), and accounted for the uncertainty due to missing data through multiple imputation. We found that residing in informal dwellings and not having post-secondary education increased the odds of HIV (aOR, 89% CrI: 1.34, 1.07–1.68 and 1.82, 1.29–2.61, respectively), after controlling for subdistrict of residence, individual, and behavioural factors. Additionally, our results suggest different pathways for how socioeconomic status (SES) affect HIV infection in males and female participants: while socioeconomic factors associated with lower SES seem to be associated with a decreased likelihood of having recently sough HIV testing among male participants, they are associated with increased sexual risk taking which, among female participants, increase the risk of HIV. Our analyses demonstrate that social determinants of health are at the root of the HIV epidemic and affect the risk of HIV in multiple ways. These findings stress the need for the deployment of programs that specifically address social determinants of health.
- Open Access
- 28 November 2022
- Published by Springer Nature in Research Square
Accelerated Gradient Methods for Sparse Statistical Learning with Nonconvex Penalties
Nesterov’s accelerated gradient (AG) is a popular technique to optimize objective functions comprising two components: a convex loss and a penalty function. While AG methods perform well for convex penalties, such as the LASSO, convergence issues may arise when it is applied to nonconvex penalties, such as SCAD. A recent proposal generalizes Nesterov’s AG method to the nonconvex setting. The proposed algorithm requires specification of several hyperparameters for its practical application. Aside from some general conditions, there is no explicit rule for selecting the hyperparameters, and how different selection can affect convergence of the algorithm. In this article, we propose a hyperparameter setting based on the complexity upper bound to accelerate convergence, and consider the application of this nonconvex AG algorithm to high-dimensional linear and logistic sparse learning problems.We further establish the rate of convergence and present a simple and useful bound to characterize our proposed optimal damping sequence. Simulation studies show that convergencecan be made, on average, considerably faster than that of the conventional proximal gradientalgorithm. Our experiments also show that the proposed method generally outperforms the current state-of-the-art methods in terms of signal recovery.
- 28 November 2022
- Published by Springer Nature in Research Square
Open AccessAccelerated Gradient Methods for Sparse Statistical Learning with Nonconvex Penalties
Nesterov’s accelerated gradient (AG) is a popular technique to optimize objective functions comprising two components: a convex loss and a penalty function. While AG methods perform well for convex penalties, such as the LASSO, convergence issues may arise when it is applied to nonconvex penalties, such as SCAD. A recent proposal generalizes Nesterov’s AG method to the nonconvex setting. The proposed algorithm requires specification of several hyperparameters for its practical application. Aside from some general conditions, there is no explicit rule for selecting the hyperparameters, and how different selection can affect convergence of the algorithm. In this article, we propose a hyperparameter setting based on the complexity upper bound to accelerate convergence, and consider the application of this nonconvex AG algorithm to high-dimensional linear and logistic sparse learning problems.We further establish the rate of convergence and present a simple and useful bound to characterize our proposed optimal damping sequence. Simulation studies show that convergencecan be made, on average, considerably faster than that of the conventional proximal gradientalgorithm. Our experiments also show that the proposed method generally outperforms the current state-of-the-art methods in terms of signal recovery.
- 24 September 2022
- Published by Elsevier in Computational Statistics & Data Analysis
- Vol. 179
A sparse additive model for high-dimensional interactions with an exposure variable
- 24 September 2022
- Published by Elsevier in Computational Statistics & Data Analysis
- Vol. 179
A sparse additive model for high-dimensional interactions with an exposure variable
- Open Access
- 1 September 2022
- Published by The R Foundation in The R Journal
- Vol. 14 (3) , 59-79
casebase: An Alternative Framework for Survival Analysis and Comparison of Event Rates
- 1 September 2022
- Published by The R Foundation in The R Journal
- Vol. 14 (3) , 59-79
Open Accesscasebase: An Alternative Framework for Survival Analysis and Comparison of Event Rates
- Open Access
- 10 May 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1-9
Novel mobility index tracks COVID-19 transmission following stay-at-home orders
Considering the emergence of SARS-CoV-2 variants and low vaccine access and uptake, minimizing human interactions remains an effective strategy to mitigate the spread of SARS-CoV-2. Using a functional principal component analysis, we created a multidimensional mobility index (MI) using six metrics compiled by SafeGraph from all counties in Illinois, Ohio, Michigan and Indiana between January 1 to December 8, 2020. Changes in mobility were defined as a time-updated 7-day rolling average. Associations between our MI and COVID-19 cases were estimated using a quasi-Poisson hierarchical generalized additive model adjusted for population density and the COVID-19 Community Vulnerability Index. Individual mobility metrics varied significantly by counties and by calendar time. More than 50% of the variability in the data was explained by the first principal component by each state, indicating good dimension reduction. While an individual metric of mobility was not associated with surges of COVID-19, our MI was independently associated with COVID-19 cases in all four states given varying time-lags. Following the expiration of stay-at-home orders, a single metric of mobility was not sensitive enough to capture the complexity of human interactions. Monitoring mobility can be an important public health tool, however, it should be modelled as a multidimensional construct.
- 10 May 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1-9
Open AccessNovel mobility index tracks COVID-19 transmission following stay-at-home orders
Considering the emergence of SARS-CoV-2 variants and low vaccine access and uptake, minimizing human interactions remains an effective strategy to mitigate the spread of SARS-CoV-2. Using a functional principal component analysis, we created a multidimensional mobility index (MI) using six metrics compiled by SafeGraph from all counties in Illinois, Ohio, Michigan and Indiana between January 1 to December 8, 2020. Changes in mobility were defined as a time-updated 7-day rolling average. Associations between our MI and COVID-19 cases were estimated using a quasi-Poisson hierarchical generalized additive model adjusted for population density and the COVID-19 Community Vulnerability Index. Individual mobility metrics varied significantly by counties and by calendar time. More than 50% of the variability in the data was explained by the first principal component by each state, indicating good dimension reduction. While an individual metric of mobility was not associated with surges of COVID-19, our MI was independently associated with COVID-19 cases in all four states given varying time-lags. Following the expiration of stay-at-home orders, a single metric of mobility was not sensitive enough to capture the complexity of human interactions. Monitoring mobility can be an important public health tool, however, it should be modelled as a multidimensional construct.
- 23 March 2022
- Published by Taylor & Francis in The American Statistician
- Vol. 76 (4) , 363-371
The “Poisson” Distribution: History, Reenactments, Adaptations
Although it is a widely used – and misused – discrete distribution, textbooks tend to give the history of the Poisson distribution short shrift, typically deriving it in the abstract as a limiting case of a binomial. The biological and physical scientists who independently derived it using space and time considerations and used it in their work are seldom mentioned. Nor are the difficulties of applying it to counts involving human activities/behavior. We (a) sketch the early history of the Poisson distribution (b) illustrate principles of the Poisson distribution involving space and time using the original biological and physical applications, as well as modern multimedia re-enactments of them, and (c) motivate count distributions accounting for extra-Poisson variation. The replayed historical applications can help today’s students, teachers and practitioners to see or hear what randomness looks or sounds like, to get practice in the practicalities of ‘counting statistics’, to distinguish situations where the pure Poisson distribution does and doesn’t hold – and to think about what one might do when it doesn’t.
- 23 March 2022
- Published by Taylor & Francis in The American Statistician
- Vol. 76 (4) , 363-371
The “Poisson” Distribution: History, Reenactments, Adaptations
Although it is a widely used – and misused – discrete distribution, textbooks tend to give the history of the Poisson distribution short shrift, typically deriving it in the abstract as a limiting case of a binomial. The biological and physical scientists who independently derived it using space and time considerations and used it in their work are seldom mentioned. Nor are the difficulties of applying it to counts involving human activities/behavior. We (a) sketch the early history of the Poisson distribution (b) illustrate principles of the Poisson distribution involving space and time using the original biological and physical applications, as well as modern multimedia re-enactments of them, and (c) motivate count distributions accounting for extra-Poisson variation. The replayed historical applications can help today’s students, teachers and practitioners to see or hear what randomness looks or sounds like, to get practice in the practicalities of ‘counting statistics’, to distinguish situations where the pure Poisson distribution does and doesn’t hold – and to think about what one might do when it doesn’t.