Max Turgeon

Projects

Mon, 01 Jan 0001 00:00:00 +0000

Here is a list of current and former projects I have worked on. Also, here is a blog post describing some of the projects I worked on as a senior biostatistician with the Saskatchewan Health Authority.

Missing data and matrix completion

Missing data is a common challenge in data science. As the number of measurements increases, so does the likelihood that at least one of them is missing for a given observation, leading to inefficient complete-case analyses. Matrix completion algorithms have gained popularity recently for their simplicity and computational efficiency. In this ongoing project, I developed a matrix completion algorithm based on generalized Singular Value Decomposition (SVD), which unlike classical SVD imposes constraints on the rows and columns of the data matrix. This framework is particularly suitable for multivariate methods like Weighted Principal Component Analysis and Correspondence Analysis.

Software

Mon, 01 Jan 0001 00:00:00 +0000

R packages

pcev: PCEV is a dimension-reduction technique, similar to Principal components Analysis (PCA), which seeks to maximize the proportion of variance (in the response vector) being explained by a set of covariates. The R package implements two estimation methods: the classical approach and a block approach, proposed by Turgeon et al. (submitted), which is suitable for high-dimensional response vectors. The package also performs inference using both analytic and permutation tests. A stable version is available on CRAN, and the development version is on Github. For more information, have a look at the vignette!
casebase: A package whose main purpose is to fit smooth-in-time parametric hazard functions using case-base sampling (Hanley & Miettinen, 2009). This approach enables the user to fit a wide class of parametric hazard functions and, as a result, to get smooth estimates of absolute risks. The website has several vignettes explaining how the package can be used.
multiKernel: This package implements multivariate prediction using kernel-machine regression. It is still very much in development on Github.
rootWishart: Functions for hypothesis testing in single and double Wishart settings, based on Roy’s largest root. This test statistic is especially useful in multivariate analysis. The computations are based on results by Chiani (2014) and Chiani (2016). They use the fact that the CDF is related to the Pfaffian of a matrix that can be computed in a finite number of iterations. This package takes advantage of the Boost and Eigen C++ libraries to perform multi-precision linear algebra. A stable version is available on CRAN, and the development version is on Github.

Talks

Mon, 01 Jan 0001 00:00:00 +0000

2025

The Statistical Frontier in Forestry and Remote Sensing
- Dalhousie University Statistics Seminar, Halifax NS
Large Scale Change Detection Using Remote Sensing Data
- Annual Meeting of the Statistical Society of Canada, Saskatoon SK
- Slides
The Role and Impact of Statistics in Industry, Public Policy, and Public Health Initiatives (panel discussion)
- Annual Meeting of the Statistical Society of Canada, Saskatoon SK

2024

AI and Landscape Management (panel discussion)
- Ontario Professional Forestry Association Conference, Sault Ste. Marie ON

2023

A Novel Approach to Stratification in the Foothills of the Sierra Nevada Mountains Using Automated Landscape Segmentation and Remote Sensed Imagery and LiDAR Data
- Society of American Foresters National Convention, Sacramento CA

2022

Generalized Soft Impute for Matrix Completion
- ICSA-Canada Chapter 2022 Symposium, Banff AB
Generalized Soft Impute for Matrix Completion
- University of Calgary Statistics Seminar, Calgary AB
- Slides
Generalized Soft Impute for Matrix Completion
- Annual Meeting of the Statistical Society of Canada, Online
- Winner of the Award for Best Presentation by a New Investigator
- Slides
What kind of training do students in Statistics and Data Science need? (panel discussion)
- Inaugural CANSSI-Prairies Summit, Winnipeg MB

2021

Investigating text data using Topological Data Analysis
- Annual Meeting of the Statistical Society of Canada, Online
Analyser les tweets de chefs politiques canadiens grâce à l’Analyse Topologique de Données
- UQAM Statistics Seminar, Université du Québec à Montréal, Montréal QC
- Slides
Investigating text data using Topological Data Analysis
- Machine Learning Special Interest Group, University of Manitoba, Winnipeg MB
- Slides

2020

Arbitrary-precision linear algebra in R using RcppEigen and BH
- useR 2020, Online conference
- Slides and Video
How to see in 100 dimensions: Transforming your data to better understand it
- Science Public Online Talks, University of Manitoba, Winnipeg MB
- Slides and Video
Principal Component of Explained Variance: High-Dimensional Estimation and Inference
- Statistics Seminar, University of Winnipeg, Winnipeg MB
- Slides

2019

Principal Component of Explained Variance: High-Dimensional Estimation and Inference
- PhD Thesis Defense, McGill University, Montreal QC
- Slides
A Tracy-Widom Empirical Estimator For Valid P-values With High-Dimensional Datasets
- ICSA-Canada Chapter 2019 Symposium, Queen’s University, Kingston, ON
- Slides
- Demo
casebase: An alternative framework for survival analysis
- Bioinformatics-Biostatistics Research Seminar, University of Manitoba, Winnipeg MB
- Slides
- Package website

2018

Principal Component of Explained Variance: High-Dimensional Estimation and Inference
- Joint Special Seminar, Departments of Statistics and Computer Science, University of Manitoba, Winnipeg MB
- Slides
Nonlinear Dimension Reduction to Improve Predictive Accuracy in Genomic and Neuroimaging Studies
- Annual Meeting of the Statistical Society of Canada, Montreal QC
- Abstract
- Slides

2017

Reduced-Rank Singular Value Decomposition for Dimension Reduction with High-Dimensional Data
- Annual Meeting of the Statistical Society of Canada, Winnipeg MB
- Winner of the 2017 Student Research Presentation Award
- Abstract
- Slides
A Novel Approach To Competing-Risk Analysis Using Case-Base Sampling
- Fifth Annual Canadian Statistics Student Conference, Winnipeg MB
- Slides

2016

Principal Component of Explained Variance: An Efficient and Optimal Data Dimension Reduction Framework for Association Studies
- Annual Meeting of the Statistical Society of Canada, St. Catherines ON
- Abstract
- Slides
A novel approach to competing risks analysis using case-base sampling
- Research Day, Department of Epidemiology, Biostatistics and Occupational Health, McGill University
- Winner of the Dr. Jim Hanley Research Day Award for Best Presentation in Biostatistics
- Slides

2015

Principal Component of Explained Variance: An Efficient and Optimal Data Dimension Reduction Framework for Association Studies
- Montreal Genomics Monthly Meeting
- Slides
Efficient dimension-reduction technique for the joint analysis of correlated phenotypes
- Annual Human and Statistical Genetics Meeting, Vancouver BC
- Slides

Teaching

Mon, 01 Jan 0001 00:00:00 +0000

I am not currently teaching any course.

Here is a list of courses I have taught in the past:

STAT 3150–Statistical Computing (Fall 2021)
- Course website
DATA 2010–Tools and Techniques in Data Science (Fall 2021)
- Course website
SCI 2000–Introduction to Data Science (Winter 2021)
- Course website
STAT 3150–Statistical Computing (Fall 2020)
- Course website
STAT 7200–Multivariate Analysis 1 (Winter 2020)
- Course website
STAT 4690–Applied Multivariate Analysis (Fall 2019)
- Course website

Guest Teaching

Dimension Reduction
- SCI 2000 Image Processing (Fall 2020), University of Manitoba, Winnipeg MB
- Slides
Regression Assumptions and Diagnostics
- CHEP 801 Advanced Epidemiology (Summer 2020), University of Saskatchewan, Saskatoon SK
- Slides
Introduction to Biostatistics
- Principles of Surgery (Winter 2019), University of Saskatchewan, Saskatoon SK
- Slides