Short Course Program

Four half-day and four full-day Short Course proposals have been selected for presentation just before the International Biometric Conference begins.  All Short Courses will take place on 8 December 2024.  These courses are taught by experienced professionals who are experts in their fields, so you do not want to miss out! 

Short Course Time Frames (Subject to change):
Full-Day Courses:  9:00 - 18:00
Morning Half-Day Courses:  9:00 - 13:00
Afternoon Half-Day Courses:  14:00 -18:00 


Full day courses

SC03 - INNOVATIVE JOINT SURVIVAL MODELS

Virginie Rondeau and Catherine Legrand

COURSE INFORMATION

Abstract
While Joint models become very common in different fields, such as medical research, literature on extending joint models outside a “classical” framework of a longitudinal biomarker and a survival time has exploded over the last decade, followed over the last years by an increased availability of softwares allowing to analyze more complex joint models.

We therefore offer in this course a broad overview over various available “innovative joint survival models” and in particular: considering joint models for recurrent events, joint models to validate surrogate endpoints, joint models for an excess of zero in the longitudinal part, joint models for a mixture of cured and uncured patients and joint models for an ordinal biomarker (with at least one survival endpoint). The objective of this course is to address with a practical perspective all these models with unified notations and level of explanation. The course will be illustrated with the analysis of several real-life datasets from medical research, considering existing R packages.

The level of the course is adapted to academic statisticians and applied statisticians in medical research (but will also be accessible to statisticians in other field of application); as well as graduate students in (bio)-statistics having already followed a course on classical survival analysis.

The two authors of this intensive course are specialist of the domain, with several international publications, books and a large experience of teaching.

    Prerequisites
    The course will provide a broad overview of various innovative joint survival models, with enough details to understand how to apply them and interpret the results, and also aim at good understanding of the estimation methods principle without entering too much in technical details.

    Therefore, the level of the course is adapted to academic statisticians and applied statisticians in medical research (but will also be accessible to statisticians in other field of application); as well as graduate students in (bio)-statistics having already followed a course on classical survival analysis.

    Participants should have (at least) some basic knowledge in the analysis of “classical” survival analysis, including in particular the concepts of (right-)censoring, the Kaplan-Meier estimator of the survival function, the logrank test and the proportional hazards model.
    Of course, the participants are expected to have some basic knowledge of statistical data analysis and inference, and in particular, already a good background on standard maximum likelihood theory and standard regression models for a continuous. This comes of course with minimum prerequisites in mathematics (matrix algebra, concept of limit, derivatives and integrals, ...). We also assume basic familiarity with the use of R in particular with regards to data import, manipulation, and standard analysis techniques for continuous and time-to-event endpoint.

    Learning Objectives
    The objective of this course is to master modern statistical methods and to master how and when applying them on
    real-world clinical data from different settings.

    This course will provide to the audience a broad overview of different up-to-date innovative extensions of the classical
    joint model and the context in which they are relevant. In particular, what to do when we have to face a non-gaussian
    longitudinal biomarker, several biomarkers or when a part of the population will never experience the event of interest
    (cure) or when we want to use a joint model to validate a surrogate markers.

    At the end of the course, the participants should be able to
    (i) recognize these situations and acknowledge the limitations of the classical approach in these situations,
    (ii) understand the important features, the estimation principle and how to apply various innovative joint models and
    make an informed choice about the different models/methods available,
    (iii) identify and use an appropriate R package to perform the analyses,
    (iv) correctly interpret the results of his/her analysis.

    For a participant whose objective is to pursue with more methodological research on one of the topic covered by the
    course, we think that this course will provide him/her with a good introduction to the topic and a good overview of the
    available estimation/fitting techniques. References for more detailed methodological descriptions of the models
    discussed will be provided.

    Textbook
    Books from the authors recommended, but not mandatory:
    1. D. Commenges, H. Jacqmin-Gadda, A. Amadou, P. Joly, B. Liquet, C. Proust-Lima, V. Rondeau, and R. Thiébaut.
    Dynamical Biostatistical Models, volume 86. CRC Press, 2015.
    2. T. Emura, S. Matsui, and V. Rondeau. Survival Analysis with Correlated Endpoints : Joint Frailty-Copula Models.
    JSS Research Series in Statistics, 2019.
    3. C. Legrand. Advanced Survival Models. Published March 23, 2021 by Chapman and Hall/CRC. ISBN
    9780367149673. 1st Edition - Catherine Legrand - Routledge (routledge.com)

    The material of this course is largely based on the content of these books.

    Laptop
    Recommended but not necessary. 

    About the Instructors

    Virginie Rondeau is the director of research in Biostatistics at the INSERM institute in Bordeaux (France) since 2015. Joint Models for recurrent events and competing risks (e.g. death) was her main research topic recently and the development of joint models for longitudinal markers and/or multiple times-to-event.

    Catherine Legrand  is Professor at the Institute of Statistics, Biostatistics and Actuarial Sciences (LIDAM/ISBA) of the Université catholique de Louvain (UCLouvain, Belgium). Her area of research includes survival data analysis with a particular focus on frailty models and cure models.

    SC04 - USING R FOR BAYESIAN SPATIAL AND SPATIO-TEMPROAL HEALTH DATA MODELING

    Andrew Lawson

    COURSE INFORMATION

    Abstract

    R is commonly use now for advanced Biostatistical applications. Bayesian spatial and spatio-temporal modeling of health data is an important topic which can be addressed using tools in R.  This course is designed for those who want to cover mapping methods, and the use of a variety of software and variants in application to small area health data.  The course will include theoretical input, covering selected  Bayesian spatial models, but also practical elements and participants will be involved in hands-on in the use of R, BRugs, Nimble,  and CARBayes  in disease mapping applications. Both human and veterinary examples will be covered in the course as well as simple space-time modelling. Examples will range over county level respiratory cancer incidence (spatial and spatio-temporal) and influenza and Covid-19 space-time modeling in South Carolina.  The course would be suitable for those with some R experience, but limited experience of spatial modeling in health applications. A recent text on this topic is

    Lawson, A. B. (2021) Using R for Bayesian Spatial and Spatio-temporal Health Modeling, CRC Press has appeared and forms the bass of this course delivery.


    Prerequisites

    • Some experience of use of R in data handling and data processing
    • Some prior Bayesian modeling experience is beneficial but not essential
    • Some spatial data analysis experience beneficial but not essential.

    Learning Objectives

    • Familiarity with Bayesian spatial health models
    • Familiarity with R use in application to spatial health data
    • Some competencies in applying Bayesian spatial models via McMC to health data problems
    • Basic familiarity with Bayesian ST modeling


    Laptop

    Recommended participants bring a laptop.

    About the Instructor
    Andrew Lawson Professor of Biostatistics in the Division of Biostatistics and Bioinformatics, Department of Public Health Sciences, College of Medicine, MUSC and is an MUSC Distinguished Professor Emeritus and ASA Fellow. His PhD was in Spatial Statistics from the University of St. Andrews, UK.

    He has over 200 journal papers on the subject of spatial epidemiology, spatial statistics and related areas. In addition to a number of book chapters, he is the author of 10 books in areas related to spatial epidemiology and health surveillance. The most recent of these is Lawson, A.B. et al (eds) (2016) Handbook of Spatial Epidemiology. CRC Press, New York, and in 2018 a 3rd edition of Bayesian Disease Mapping; hierarchical modeling in spatial epidemiology  CRC Press. In 2021. a new volume entitled Using R Bayesian Spatial and  Spatio-temporal Health modeling CRC Press appeared. He has acted as an advisor in disease mapping and risk assessment for the World Health Organization (WHO) and is founding editor of the Elsevier journal Spatial and Spatio-temporal Epidemiology.  Dr Lawson has delivered many short courses in different locations over the last 20 years on Bayesian Disease Mapping with OpenBUGS, INLA, and Nimble, and more general spatial epidemiology topics.
    Web site: http://people.musc.edu/~abl6/ 

    Half day courses

    SC05 - Causal inference in drug development: Applications and methods

    Robin Dunn, Jiarui Lu, and Frank Bretz

    COURSE INFORMATION

    Abstract

    Causal thinking and related inference methods are gaining increasing prominence in global drug development in light of the addendum to the E9 guideline on 'Statistical principles in clinical trials' by the International Council of Harmonization (ICH, 2019) and the guideline on covariate adjustment by the U.S. Food and Drug Administration (FDA, 2023). These guidelines refer to terminology, concepts and methods from the causal inference literature, such as potential outcomes, principal stratification, non-collapsibility and standardization.

    Although widely underutilized in drug development, causal inference methods can add value in many settings, including those outlined in the two guidelines above, but also for the use of external control data through the target trial framework and understanding cause and effect in pharmacometric and pharmacovigilance applications. In this course, we focus on causal inference methods tailored to the challenges commonly encountered in randomized controlled trials occurring in drug development. We start with an introduction to the ICH E9(R1) estimand framework and basic causal inference concepts, followed by a detailed discussion of causal inference methods targeting hypothetical estimands, as well as conditional vs. marginal treatment effects. We illustrate the methods with case studies and provide code examples to facilitate implementation in practice.

     

    Prerequisites

    The participants should have basic knowledge of the fundamentals of statistics including experience with common data types (continuous, binary, time-to-event) as well as the associated models and estimation methods, such as maximum likelihood, general linear models, and Cox models. Knowledge in causal inference is a plus, but not mandated. Moreover, participants are expected to have basic knowledge of clinical trial methodology and should be familiar with concepts and terms such as bias, randomization, and blinding.

    The difficulty level of the course is intermediate, at a second-year graduate course level. The focus will not be on the theoretical derivations and properties of the statistical methods, but on their application in clinical trial settings.

    Learning Objectives

    The difficulty level of the course is intermediate, at a second-year graduate course level. The learning objectives are four-fold:

    1. to describe basic concepts of causal inference (e.g., potential outcomes, main assumptions, confounders) and appreciate the role of causal inference in randomized clinical trials;
    2. to identify and apply common estimation methods of causal effects relevant to clinical trials in drug development;
    3. to implement appropriate analyses in practical settings; and
    4. to get an overview of basic functionality in R to design and analyze clinical trials.

    About the Instructors

    Dr. Robin Dunn is a Senior Principal Statistical Consultant in the Advanced Exploratory Analytics group at Novartis. She has worked on causal inference research and consulting projects at Novartis involving topics such as external controls to support a small placebo arm, covariate adjustment for efficient inference, and heterogeneous dose effects. She has also been involved in developing and presenting Novartis trainings on causal inference in both randomized controlled trials and trial-external data settings. Prior to joining Novartis, she received her PhD in Statistics from Carnegie Mellon University in 2021.

    Dr. Jiarui Lu is a Principal Biostatistician in the Biostatistics Department at Vertex Pharmaceuticals, Boston, MA, USA. His has been supporting multiple clinical and research projects on causal inference and estimand. His research interests include causal inference, estimands, treatment effect heterogeneity, and real world evidence. He is particularly interested in implementing and applying these innovative methods to drug development. Before joining Vertex, he was a statistical consultant at Novartis, and he got his Ph.D. in Biostatistics from the University of Pennsylvania.

    Prof. Frank Bretz is a Distinguished Quantitative Research Scientist at Novartis. He has supported the methodological development in various areas of biostatistics, including dose finding, multiple comparisons, and adaptive designs. Frank is currently holding Adjunct professorial positions at the Hannover Medical School (Germany) and the Medical University of Vienna (Austria). He was a core member of the ICH E9(R1) working group on “Estimands and sensitivity analysis in clinical trials”. Frank was an Executive Board member of the International Biometric Society (IBS) and served as the President of the IBS Austro-Swiss Region (IBS-ROeS). He is a recipient of the Susanne-Dahms-Medal from the IBS German Region (IBS-DR) and a Fellow of the American Statistical Association.

    SC06 - Statistical and machine learning for big geospatial data

    Dr. Abhirup Datta

    COURSE INFORMATION

    Abstract
    Geospatial data, routinely encountered in environmental health, climate sciences, disease epidemiology, forestry, and ecology, have traditionally been analyzed using statistical models built on foundations of stochastic processes. Increasingly, practitioners are adopting machine learning methods for geospatial analysis. Should the decades of development in spatial statistics be abandoned for black-box machine learning? This short course demonstrates the pitfalls of naive machine learning for geospatial data and offers a tutorial on state-of-the-art hybrid machine learning methods for geospatial data that leverage well-established spatial statistics principles.

    We review traditional geospatial approaches like the linear mixed model using Gaussian process (GP), and non-linear machine learning methods like random forests and neural networks. We then present hybrid approaches that embed non-linear machine learning within traditional geospatial mixed effect models, relaxing the stringent assumption of linearity, while preserving Gaussian process random effects and thereby retaining the interpretability, flexibility, and parsimony for estimation and prediction. We present recent methods with different choices of machine learning algorithms (random forests and neural networks) and different data types (continuous and binary), provide demonstrations with published software on real and simulated data. The tutorial will equip practitioners with a suite of hybrid machine learning methods and software tailored for geospatial analysis.

    Prerequisites
    The course will cover topics of various levels of difficulty and is intended for audience with diverse quantitative backgrounds. Much of the course will focus on application of the new hybrid geospatial machine learning methods using published software and live demonstration using data examples. This content will be accessible to practitioners from a wide range of fields (environmental health, climate sciences, disease epidemiology, forestry, ecology) interested in learning modern tools for geospatial analysis. A thorough review will be provided to introduce the audience to both geostatistics as well as popular machine learning methods. Some of the advanced materials, particularly, methodological, and computational details of the new hybrid machine learning algorithms, will be taught at the level of a graduate course for students in statistics, biostatistics, computer science or related fields.

    While the course will primarily focus upon practical modeling, computing and data analysis, short course participants will benefit from some prior understanding of mathematical statistics and linear algebra at the undergraduate or advanced undergraduate level. We will not assume any significant previous exposure to geospatial methods or machine learning algorithms, although students with basic knowledge of the area will certainly face a gentler learning curve. All the computational tools and environments will also be introduced as necessary in the course, but some experience with the R-programming language for statistical analysis will be helpful. Experience with GIS or any other specialized software is not required.

    Learning Objectives
    From this course, the participants will acquire:
    1. understanding of the basics of geospatial data analysis and visualization
    2. understanding the strengths of spatial linear mixed effect models (interpretability, predictive capability) and their limitations (assumption of linearity)
    3. familiarity with popular non-linear machine learning algorithms like random forests and neural networks and the perils of naive applications of these methods for spatially correlated data
    4. understanding state-of-the-art hybrid methods that embeds non-linear machine learning within spatial mixed models (random forests (RF-GLS) and neural networks (NN-GLS))
    5. familiarity with newly developed software for analyzing large geospatial datasets with RF-GLS and NN-GLS for different types of geospatial data

    About the Instructor

    Dr. Abhi Datta is Associate Professor of Biostatistics at the Johns Hopkins Bloomberg School of Public Health (JHSPH). He received his PhD in Biostatistics from the University of Minnesota in 2016 and was an Assistant Professor in the Department of Biostatistics at JHSPH from 2016-2021. Dr. Datta have published 45 peer-reviewed articles (23 as first or senior author). His publications have appeared in prestigious journals, including the Journal of the American Statistical Association, the Annals of Statistics, the Annals of Applied Statistics, Biometrika, Biometrics, Biostatistics, Atmospheric Environment, and the Proceedings of the National Academy of Sciences. His research as principal investigator has been supported by grants from the National Science Foundation (NSF), the Bill and Melinda Gates Foundation (BMGF), and the National Institutes of Environmental Health Sciences (NIEHS R01). Dr. Datta’s research has been recognized via multiple national and international awards including the Young Statistical Scientist Award (YSSA) by the International Indian Statistical Association, Abdel El-Shaarawi Early Investigator's Award by The International Environmetrics Society, and Early Investigator Award from the American Statistical Association Section on the Environment.

    SC07 - A practical course in difference-in-differences

    Laura Hatfield

    COURSE INFORMATION

    Abstract
    This course aims to provide participants with a solid grounding in methods for Difference-in-Differences (DID), a popular method for quasi-experimental causal inference in the social sciences.

    The course begins with a potential outcomes approach to constructing target estimands. Various approaches to construct the counterfactual will be discussed. Then we will explore DID methods in depth, including the required causal assumptions. Selecting appropriate comparison groups is crucial for ensuring the plausibility of these assumptions. The course will provide strategies for identifying suitable comparison groups, with special attention to matching, weighting, and regression approaches.

    Participants will learn how to align the estimation method with the target estimand. Analyses for staggered adoption will be discussed. We will also discuss inference, especially for small numbers of clusters and highlight approaches such as aggregation and permutation.
    For sensitivity analyses, we will cover non-inferiority/equivalence tests as an alternative to conventional parallel trends testing, placebo tests, event study plots, negatively correlated control groups, and worst-case differential trends. Relate methods such as synthetic controls, lagged dependent variables, and remixes of existing techniques, will also be introduced. The course concludes with a literature round-up, highlighting new developments and useful reviews.

    Prerequisites
    Attendees should be familiar with statistics on the level of a year-long graduate-level course in statistical
    modeling/inference, especially regression-based estimation and inference. Familiarity with basic concepts of causal
    inference (e.g., potential outcomes, confounding, target estimands, identification assumptions) will also be helpful. The
    course includes no programming, but the tools recommended to implement the techniques discussed in the course will
    primarily be in R.

    Learning Objectives
    This course is intended for researchers who want to be responsible users of difference-in-differences methods but do not have time to keep up with the deluge of new methods developments. At the end of the course, participants will be able to
    1. Formally define the target estimand and the required causal assumptions of a difference-in-differences study
    2. Choose plausible comparison groups and address potential confounding
    3. Estimate the causal target using methods that are compatible with the causal assumptions
    4. Perform statistical inference using flexible methods
    5. Conduct principled sensitivity analyses
    6. Be familiar with recent methodological developments

    Laptop
    Not needed.

    About the Instructor
    Laura Hatfield, PhD, is an associate professor of health care policy (biostatistics) in the Department of Health Care Policy at Harvard Medical School. Her methods research focuses on causal inference in non-randomized settings, especially using difference-in-differences, and quantifying variation in health care utilization, outcomes, and quality using clustering and hierarchical Bayesian models. Hatfield earned her BS in genetics from Iowa State University and her MS and PhD in biostatistics from the University of Minnesota. Dr. Hatfield developed and taught this short course for the Harvard Catalyst program in Feb 2023; more than 430 people registered for that course.