ENAR Webinar Series (WebENARs)

Past Webinars

 

Approximate Bayesian model selection as an alternative to classical hypothesis testing: Writing Outside of the Statistical Literature

Thursday, October 26, 2023
1-3 pm

Presenter:
Christopher T. Franck, Virginia Tech

Abstract:
By now, statisticians and the broader research community are aware of the controversies surrounding traditional hypothesis testing and p-values. Many alternative viewpoints and methods have been proposed, as exemplified by The American Statistician's recent special issue themed "World beyond p<0.05." However, it seems clear that the broader scientific effort may benefit if alternatives to classical hypothesis testing are described in venues beyond the statistical literature. This paper addresses two relevant gaps in statistical practice. First, we describe three principles statisticians and their collaborators can use to publish about alternatives to classical hypothesis testing in the literature outside of statistics. Second, we describe an existing BIC-based approximation to Bayesian model selection as a complete alternative approach to classical hypothesis testing. This approach is easy to conduct and interpret, even for analysts who do not have fully Bayesian expertise in analyzing data. Perhaps surprisingly, it does not appear that the BIC approximation has yet been described in the context of "World beyond p<0.05." We address both gaps by describing a recent collaborative effort where we used the BIC-based techniques to publish a paper about hypothesis testing alternatives in a high-end biomechanical engineering journal.

Bio:

Purchase Webinar Recording (10/26/23).

 

Leading with Impact: Enhancing Effective Leadership in Biostatistics Research Teams

Friday, November 3, 2023
1-3 pm

Presenters:
Stefanie Robel, University of Alabama at Birmingham
Kendra Sewall, Virginia Tech
Shwetal Mehta, Barrow Neurological Institute

Abstract:

Running an impactful biostatistics research program involves more than a solid plan for developing and applying mathematical and statistical techniques. Effective leadership and team management are critical yet frequently overlooked components, essential for research project success. Biostatisticians leading or part of a team often face diverse and challenging personalities, different motivation levels, and varying expertise, leading to potential conflict, miscommunication, and decreased productivity.

"Leading with Impact" is a webinar tailored for biostatisticians aiming to bolster their leadership skills. The session begins with a presentation that comprehensively covers three critical leadership aspects:

• Understanding when to apply various leadership styles,
• Balancing research goals with trainee career objectives, and
• Addressing the issue of unmotivated team members.

Participants will gain practical strategies for navigating these challenges, enhancing team performance and satisfaction.

The session continues with a panel discussion featuring Drs. Kendra Sewall, Stefanie Robel, and Shwetal Mehta. All panelists actively run research programs, hold administrative leadership positions, or act as leadership coaches. They will share their real-world experiences and practical solutions and address audience questions, further empowering attendees to emerge as confident and effective team leaders.

Bios:

 

 

 

Purchase Webinar Recording (11/3/23).

 

Statistical Challenges in Clinical Trials for COVID-19

Friday, June 2, 2023
10 a.m. to 12 noon

One Uncertainty Too Many? Managing Unknowns in Clinical Trials of Outbreak Diseases
The West African Ebola virus disease outbreak taught us the importance of starting clinical trials rapidly, to find effective intervention before the outbreak ends, while interventions may be useful to reduce disease burden and suffering. However, during novel infectious disease outbreaks, there are often no precedents for trial design and limited data exist to design an optimal trial. Waiting for more certainty about key design characteristics may delay trial implementation too significantly, creating a dilemma. Start now with an imperfect design or delay study start until the design is better defined? In this talk, I will review my experiences from designing studies for Ebola and COVID-19 treatment studies and emphasize the importance of incorporating sufficient design flexibility without compromising scientific rigor.

 

Issues Encountered While Monitoring the US Government-Supported Covid-19 Vaccine Trials
Operation Warp Speed (OWS) was a partnership among vaccine companies, government agencies and academia created during the Covid-19 global pandemic. A primary goal was to accelerate the development of Covid-19 vaccines. A fundamental principle that the OWS program established was that the National Institutes of Health would oversee a single Data and Safety Monitoring Board (DSMB) to review and monitor all OWS vaccine trials. This was implemented by having a statistician from NIAID be the executive secretary for the Board. The formation of OWS and its requirement for a single DSMB played an important role in the rapid development and delivery of effective Covid-19 vaccines. In this talk I will describe the unique issues and challenges faced while monitoring these trials and provide suggestions for future similar endeavors.

 

Multiplicity Issues in Platform Trials of COVID-19
The COVID-19 pandemic led to platform trials comparing several active arms to a common control arm. An important question is whether to make a multiple comparison adjustment, as is traditionally done. Pressure to find an effective treatment quickly for a dangerous disease argues against such an adjustment, but statistical concerns seem to support an adjustment. We will take a close look at the statistical arguments and give both sides of the debate. Then we will consider a multiple comparison issue arising from the first trial to show benefit of an intervention in COVID-19, namely the Adaptive Coronavirus Treatment Trial (ACTT-1) of Remdesivir versus placebo. There was a statistically significant effect on the primary endpoint of time to recovery and on an important secondary endpoint of the World Health Organization’s 8-point ordinal scale at Day 15. Results for mortality did not quite reach statistical significance, but there was an apparent effect in ordinal scale 5, patients requiring supplemental oxygen (not including high-flow or invasive mechanical ventilation). Is the apparent effect of remdesivir on mortality in OS-5 real, or is it just the play of chance? If it is real, could it be part of an overall benefit of Remdesivir on mortality, or is it OS-5-specific? We will address these important questions.

 

Powerful Data Presentations with Success = (PD)²

Friday, April 28, 2023
1 p.m. to 3 p.m.

Abstract:
The amount of available data has exploded in the past ten years. Those holding quantitative roles in academia, government, and industry have been called upon to analyze and interpret it. We must turn data into information so that researchers can advance science and decision makers can act upon these advances. This is true whether you are assessing the performance of a new therapy, the impact of public health policy, or modeling disease transmission. It is also true for those presenting new statistical analysis methods. The point of a presentation is to synthesize facts in a meaningful, digestible way for the audience. Yet, many presentations serve to create confusion and because they lack clarity. Even worse, the lack of clarity leads to poor decisions built on erroneous interpretations.

Developing good communication skills around scientific and statistical presentations is essential for those in quantitative fields. This talk will discuss a four phased process for excelling at these types of presentations. The four phases are Prepare, Design, Practice and Deliver or (PD)2 for short. In the context of the process, emphasis will be placed on explaining complex concepts, formatting your results for clarity, designing your slides to facilitate interpretation, and engaging a non-quantitative or mixed level audience. The material will also cover incorporating story into a scientific presentation and effective principles of data visualization. You will leave the talk with an overall framework for tackling your next formal presentation as well as tips and tricks you can immediately use in informal team interactions.

Purchase Webinar Recording (4/28/23).

 

Statistical Issues in Responsible Conduct of Research

Friday, March 3, 2023
10 a.m. to 12 p.m.

Abstract:
Technology has made it easier to share and analyze data. As statisticians, we are responsible for the data that is in our “custody,” and how it is used. This WebENAR will discuss some of the ethical issues that arise in collaborative research, the importance of reproducible analyses, and the impact that both can have on “good science.”

Purchase Webinar Recording (3/3/23).

 

Rethinking Race-ethnicity: Introducing Novel Survey-based Measures of Lifetime Experience of Discrimination and Stress

Friday, November 4, 2022
2 to 3 p.m. Eastern

Abstract:
Though race-ethnicity is not a biological variable, race-ethnicity is included in nearly every medical study and often very statistically & meaningfully significant. New measures are critically needed that will allow biomedical researchers to disentangle race-ethnicity from the true individual, interpersonal, and structural causes of health disparities. In this talk, we introduce novel measures using simple survey items to capture self-reported experience of discrimination and stress spanning the life course agnostic to source. Early results from an employee survey will presented.

Purchase Webinar Recording (11/4/22).

 

Collab: Heterogeneous Causal Effects of Neighborhood Policing in New York City with Staggered Adoption of the Policy and Evaluating Methods to Estimate the Effect of State Laws on Opioid-related Outcomes in the Presence of Confounding

Friday, October 21. 2022
10 a.m. to 12 p.m. Eastern

Abstract:
In New York City, neighborhood policing was adopted at the police precinct level over the years 2015-2018, and it is of interest to both (1) evaluate the impact of the policy, and (2) understand what types of communities are most impacted by the policy, raising questions of heterogeneous treatment effects. We develop novel statistical approaches that are robust to unmeasured confounding bias to study the causal effect of policies implemented at the community level. We find that neighborhood policing decreases discretionary arrests in certain areas of the city, but has little effect on crime or racial disparities in arrest rates.

 

Abstract:
The nation is in the midst of an opioid-related public health crisis. In response, states have enacted a heterogeneous collection of policies aimed at reducing mortality and morbidity, producing a state policy landscape that is complex and dynamic. Understanding how best to estimate policy effects is important and several unanswered questions remain, particularly about optimal methods for handling confounding bias that will result when states implementing a policy are different from states that do not. Using simulations, we examined the statistical properties of several statistical methods to estimate the effects of state-level opioid policies to empirically identify the best methods for handling confounding. Findings from these simulations can help identify which models are best to robustly estimate state-level policy effects on opioid-related outcomes. Identifying robust and powerful methods are needed to help ensure future policy decisions are based on results from well-designed evaluations that yield accurate policy effects.

Purchase Webinar Recording (10/21/22).

 

Recent Developments in the Analysis of Adaptive Designs and Their Relevance to Platform Trials

Friday, June 3, 2022
10 a.m. to 12 p.m.

 

Combining the Three Cultures of Quantitative Decision Making in Drug Development

Friday, May 20, 2022
1-3 pm EST

Purchase Webinar Recording (5/20/22).

 

How Early Career Survival Skills Often Turn Into Mid-career Bottlenecks: A Researcher’s Journey Perspective

Friday, May 6, 2022
1-3 pm EST

Purchase Webinar Recording (5/6/22).

 

A Novel Approach for the Analysis of Randomized Clinical Trials

Friday, April 22, 2022
10 a.m. to 12 p.m.

Purchase Webinar Recording (4/22/22).

 

Collaboration: Pairwise Survival Analysis and Causal Inference for Infectious Disease Epidemiology and Understanding Transmission Dynamics of Emerging Infectious Diseases from Contact-tracing Data

Friday, February 25, 2022
10 a.m. to 12 p.m. ET

Purchase Webinar Recording (2/25/22).

 

Estimands, Estimators, and Estimates: Aligning Target of Estimation, Method of Estimation, and Sensitivity Analysis, with Application to the COVID-19 Pandemic

Friday, November 19, 2021
10 a.m. to 12 p.m. ET

Abstract:
The ICH E9(R1) Addendum on 'Estimands and Sensitivity Analysis in Clinical Trials' introduced a framework to align planning, design, conduct, analysis, and interpretation of clinical trials. When defining the clinical question of interest, clarity is needed about 'intercurrent events' that affect either the interpretation or the existence of the measurements associated with the clinical question of interest, such as discontinuation of assigned treatment, use of an additional or alternative treatment and terminal events such as death. The description of an estimand should reflect the clinical question of interest in respect of these intercurrent events, and the Addendum introduces strategies to reflect different questions of interest that might be posed. The choice of strategies can influence how more conventional attributes of a trial are reflected when describing the clinical question, for example the treatments, population or the variable (endpoint) of interest.

In this seminar we briefly introduce the estimand framework according to the ICH E9(R1) Addendum and describe various strategies for addressing intercurrent events when defining the clinical question of interest. We then reflect on the experience and lessons learned of implementing the Addendum through an internal cross-functional and cross-divisional working group that encompasses various estimand initiatives. Next, we discuss in detail the hypothetical estimand strategy, where a scenario is envisaged in which the intercurrent event would not occur. The Addendum acknowledges that a wide variety of hypothetical scenarios can be envisaged, but it also clarifies that some scenarios are likely to be of more clinical or regulatory interest than others. We share our experiences and try to provide some guidance on their use in clinical trial practice. Finally, we demonstrate how the estimand framework can usefully be applied to clinical trials impacted by the COVID-19 pandemic to address potential pandemic-related trial disruptions and embed them in the context of study objectives and design elements. We introduce different hypothetical estimand strategies and review various causal inference and missing data methods such as multiple imputation and (augmented) inverse probability weighting for the estimation step. To clarify, we describe the features of a stylized trial in neuroscience, and how it may have been impacted by the pandemic. This stylized trial will then be re-visited by discussing the changes to the estimand and the estimator to account for pandemic disruptions.

Purchase Webinar Recording (11/19/21).

 

Novel Applications of Real-world Data to Support Clinical Trials

Friday, June 11, 2021
1 p.m. to 3 p.m. ET

Abstract:
Real-world data (RWD) have played an increasingly important role in healthcare decisions, for example supporting the design, analysis, and contextualization of clinical trials. In this webinar, we invite panelists from industry with backgrounds in statistics, clinical trials, data science, epidemiology, and health outcomes research to discuss novel applications where RWD can be used to support clinical trials.

We discuss propensity-score based methods to leverage RWD as an external data source to augment single-arm clinical trials, enhanced extrapolation of long-term clinical trial survival outcomes using electronic health record (EHR)-derived RWD, and enrollment projection for a prospective pragmatic trial design that utilizes RWD. Finally, we tie the three topics together with a panel discussion.

Purchase Webinar Recording (06/11/21)

 

Incorporating Diversity, Equity and Inclusion in Biostatistics Courses

Friday, April 23, 2021
10 a.m. to 12 p.m. Eastern

Abstract:
As we embrace conversations about improving diversity, equity, and inclusion (DEI) in the field of biostatistics, ideally, these perspectives should appear in every aspect of the profession, including incorporating these principles into how we teach our trainees. By incorporating DEI into biostatistics pedagogy, instructors and trainees can cultivate a more holistic understanding of both historical background and current challenges in the field by enabling all students to see themselves in the content and how they might contribute to making important contributions to both statistical theory and application.

We will begin this WebENAR by putting this into historical context to establish the importance of incorporating DEI into biostatistics training and coursework. We will then introduce practical examples from our own experiences of how to introduce these concepts into courses without compromising course objectives and without requiring additional time for these modifications. The session will conclude with open discussion where we encourage all those attending the WebENAR to share their own experiences and ideas for making biostatistics courses more diverse, equitable, and inclusive.

Purchase Webinar Recording (04/23/21)

 

Slamming the Sham: A Bayesian Model for Adaptive Adjustment with Noisy Control Data

Friday, March 26, 2021
10 a.m. to 12 p.m. ET

Abstract:
It is not always clear how to adjust for control data in causal inference, balancing the goals of reducing bias and variance. We show how, in a setting with repeated experiments, Bayesian hierarchical modeling yields an adaptive procedure that uses the data to determine how much adjustment to perform. The result is a novel analysis with increased statistical efficiency compared to the default analysis based on difference estimates. We demonstrate this procedure on two real examples, as well as on a series of simulated datasets. We show that the increased efficiency can have real-world consequences in terms of the conclusions that can be drawn from the experiments. We also discuss the relevance of this work to causal inference and statistical design and analysis more generally. This is joint work with Matthijs Vákár. Before attending the talk, people are encouraged to read our paper: http://www.stat.columbia.edu/~gelman/research/unpublished/chickens.pdf

Purchase Webinar Recording (03/26/21)

 

Revisiting ICH E9 (R1) During the COVID-19 Pandemic

Friday, January 22, 2021
10 a.m. to 12 p.m. ET

Presenters:
Yongming Qu, Eli Lilly
Yongming Qu is currently a Sr. Research Fellow at Eli Lilly and Company. He received his PhD in Statistics from Iowa State University. He has provided key leadership in various stages drug clinical development at Lilly. He has been passionate in developing new statistical methods for better clinical trial design and data analysis that impact drug development. He published more than 70 articles in statistical, medical and mathematical journals, and is an ASA Fellow.

Ilya Lipkovich, Eli Lilly
Ilya Lipkovich is a Sr. Research Advisor at Eli Lilly and Company. Ilya received his PhD in Statistics from Virginia Tech in 2002 and has more than 15 years of statistical consulting experience in pharmaceutical industry. He is an ASA Fellow and published on subgroup identification in clinical data, analysis with missing data, and causal inference. He is a frequent presenter at conferences, a co-developer of subgroup identification methods, and a co-author of the books "Analyzing Longitudinal Clinical Trial Data. A Practical Guide" and "Estimands, Estimators and Sensitivity Analysis in Clinical Trials."

Abstract The current COVID-19 pandemic poses numerous challenges for ongoing clinical trials and provides a stress-testing environment for the existing principles and practice of estimands in clinical trials. The pandemic may increase the rate of intercurrent events (ICEs) and missing values, spurring a great deal of discussion on amending protocols and statistical analysis plans to address these issues. In this article we revisit recent research on estimands and handling of missing values, especially the ICH E9 (R1) on Estimands and Sensitivity Analysis in Clinical Trials. Based on an in-depth discussion of the strategies for handling ICEs using a causal inference framework, we suggest some improvements in applying the estimand and estimation framework in ICH E9 (R1). Specifically, we discuss a mix of strategies allowing us to handle ICEs differentially based on the causes of ICEs. We also suggest ICEs should be handled primarily by hypothetical strategies and provide examples of different hypothetical strategies for different types of ICEs as well as a road map for estimation and sensitivity analyses. We conclude that the proposed framework helps streamline translating clinical objectives into targets of statistical inference and resolves many issues with defining estimands and choosing estimation procedures arising from unanticipated events such as the current pandemic.

Purchase Webinar Recording (01/22/21)

 

Role of Statisticians in a Pandemic

Friday, November 13, 2020
10 a.m. to 11:30 a.m. Eastern

Moderator:

Panelists:

 

 

 

Abstract:
While the topic is very broad, we shall try to: (1) highlight some specific unique challenges based on the nature of the pandemic, e.g. our lack of knowledge about the virus coming in, the urgency to learn act quickly, yet the necessity to think careful and rigorously to avoid false steps and conclusions. (2) clearly communicate the importance of our profession and people with our quantitative skill sets to engage and have a seat at the table to have our perspective heard, both by policymakers and the media, during this crisis.

People in our profession need to have better communication with policymakers, and many in our field might not recognize their potential or the importance of our skillset and perspective to the big decisions going on in society. We hope our panel discussion can inspire more statisticians to get engaged in this way.

Purchase Webinar Recording (11/13/20)

 

(Almost) All of Entity Resolution

October 2, 2020
10 a.m. to 12 p.m. Eastern

Abstract:
Whether the goal is to estimate the number of people that live in a congressional district, to estimate the number of individuals that have died in an armed conflict, or to disambiguate individual authors using bibliographic data, all these applications have a common theme - integrating information from multiple sources. Before such questions can be answered, databases must be cleaned and integrated in a systematic and accurate way, commonly known as record linkage, de-duplication, or entity resolution. In this article, we review motivational applications and seminal papers that have led to the growth of this area. Specifically, we review the foundational work that began in the 1940's and 50's that have led to modern probabilistic record linkage. We review clustering approaches to entity resolution, semi- and fully supervised methods, and canonicalization, which are being used throughout industry and academia in applications such as human rights, official statistics, medicine, citation networks, among others. Finally, we discuss current research topics of practical importance.

Purchase Webinar Recording (10/02/20)

 

The Role of Statistics in Transforming EHR Data into Knowledge

Friday, June 19, 2020
10 a.m. to 12 p.m. Eastern

Abstract:
The widespread adoption of electronic health records (EHR) as a means of documenting medical care has created a vast resource for research on health conditions, interventions, and outcomes. Informaticians have played a leading role in the process of extracting “real world data” from EHR, with statisticians playing a more peripheral part. However, statistical insights on study design and inference are key to drawing valid conclusions from this messy and incomplete data source. This webinar will describe the basic structure of EHR data, highlight key challenges to research arising from this data structure, and present an overview of some statistical methods that address these challenges. The discussion of issues related to the structure and quality of EHR data will include: data types and methods for extracting variables of interest; sources of missing data; error in covariates and outcomes extracted from EHR and claims data; and data capture considerations such as informative visit processes and medical records coding procedures. The overall goal of this webinar is to illustrate the unique contribution of statistics to the process of generating knowledge from EHR data and equip participants with some tools for doing so.

Purchase Webinar Recording (6/19/20)

 

Spatial Statistics for Disease Ecology

Friday, April 17. 2020
10 a.m. to 12 p.m. Eastern

Abstract:
The field of disease ecology involves exploration of the multiple, dynamic interactions between pathogens, hosts, and the environment that result in the transmission of disease. Many of these interactions involve spatial or spatiotemporal components that determine the course of an outbreak and may offer potential interventions to stem the extent and duration of an outbreak at the population level. In this webinar, we provide a brief overview of the field of disease ecology, the motivating questions of interest, the nature of data involved, and the interaction of statistical and mathematical modeling addressing these questions with available data. We offer two illustrations relating to the monitoring and analysis of zoonic disease, namely: identifying geographic drivers of the spread of raccoon rabies along the Eastern United States and utilizing environmental data to enhance animal surveillance for plague in California. Both examples utilize concepts and analytic tools from spatial statistics to better understand and monitor geographically-referenced zoonotic diseases in wild animal populations.

Purchase Webinar Recording (4/17/20)

 

Collaboration: Applications of RWE in drug development and methodologies for confounding control and A statistical roadmap for journey from real-world data to real-world evidence

Friday, February 28. 2020
10 a.m. to 12 p.m. Eastern

Abstract:
Applications of RWE in Drug Development and Methodologies for Confounding Control: Real-world evidence (RWE) is playing an increasingly important role in drug development, from early in discovery throughout clinical development program to life-cycle management. RWE can augment randomized clinical trials for regulatory approval, establish the effectiveness and safety profile in routine clinical practice to support reimbursement decision, and constitute an integral part of scientific communication overall. Due to its noninterventional nature, a key challenge of robust RWE generation is to establish causal relationship between exposure and outcome. This talk focuses on several main methodologies for causal inference, consisting of IPTW, MLE, AIPTW, and TMLE using full data and matched data that is derived from propensity score matching, respectively. Following the RWE roadmap outlined in the first talk, practical considerations are given to facilitate the series of decisions for confounding control, such as defining estimand, usage of matching, and choice among different analytic frameworks.

 

Abstract:
A statistical roadmap for journey from real-world data to real-world evidence: Randomized controlled clinical trials (RCTs) are the gold standard for evaluating the safety and efficacy of pharmaceutical drugs, but in many cases their costs, duration, limited generalizability, and ethical or technical feasibility have caused some to look for real-world studies as alternatives. On the other hand, real-world data may be much less convincing due to the lack of randomization and the presence of confounding bias. In this presentation, we propose a statistical roadmap to translate real-world data (RWD) to robust real-world evidence (RWE). The roadmap consists of three main stations: (1) defining an estimand translating the research objective into a precise definition of the treatment effect that is to be estimated, (2) constructing an efficient estimator (minimum-variance unbiased estimator) for the estimation of the estimand, and (3) conducting sensitivity analysis to explore the robustness of the inference to deviation from the underlying no-unmeasured confounding assumption. The Food and Drug Administration (FDA) is working on guidelines, with a target to release a draft by 2021, to harmonize RWD applications and monitor the safety and effectiveness of pharmaceutical drugs using RWE. The proposed roadmap aligns with the newly released framework for FDA's RWE Program in December 2018 and we hope this statistical roadmap is useful for statisticians who are eager to embark on their journeys in the real-world research.

Purchase Webinar Recording (2/28/20)

 

The Central Role of Personalized Solution in the Era of Digital Health

Friday, January 10, 2020
10 a.m. to 12 p.m. Eastern

Abstract:
Digital health is an important pharmaceutical industry trend in recent years, and it can bring significant disruptive innovation to transform healthcare industry. In this talk, we will provide an introduction on digital health and associated analytics challenges and opportunities. In particular, we will focus on the central role of personalized intervention in the era of digital health.

Purchase Webinar Recording (1/10/20)

 

New Statistical Learning Methods for Optimizing Dynamic Treatment Decision Rules Leading Toward Personalized Health Care

Friday, December 6, 2019
10 a.m. to 12 p.m. Eastern

Abstract:
In this talk, we present recent advances and statistical developments for evaluating Dynamic Treatment Regimes (DTR), which allow the treatment to be dynamically tailored according to evolving subject-level data. Identification of an optimal DTR is a key component for precision medicine and personalized health care. Specific topics covered in this talk include several recent projects with robust and flexible methods developed for the above research area. We will first introduce a dynamic statistical learning method, adaptive contrast weighted learning (ACWL), which combines doubly robust semiparametric regression estimators with flexible machine learning methods. We will further develop a tree-based reinforcement learning (T-RL) method, which builds an unsupervised decision tree that maintains the nature of batch-mode reinforcement learning. Unlike ACWL, T-RL handles the optimization problem with multiple treatment comparisons directly through a purity measure constructed with augmented inverse probability weighted estimators. T-RL is robust, efficient and easy to interpret for the identification of optimal DTRs. However, ACWL seems more robust against tree-type misspecification than T-RL when the true optimal DTR is non-tree-type. At the end of this talk, we will also present a new Stochastic-Tree Search method called ST-RL for evaluating optimal DTRs.

Purchase Webinar Recording (12/6/19)

 

Subgroup Identification for Differential Treatment Effects

Friday, November 8, 2019
10 a.m. to 12 p.m. Eastern

Abstract:
Many subgroup identification methods exist but they have not been compared together. To better understand the relative strengths and weaknesses of the methods, we briefly review those with publicly available software (FindIt, GUIDE, Interaction Trees, MOB, Outcome Weighted Estimation, PRIM, ROWSi, Sequential Batting, SIDES, and Virtual Twins) and then compare their performance on seven criteria: (i) variable selection bias, (ii) probability of false discovery, (iii) probability of correct variable identification, (iv) bias in subgroup treatment effect estimates, (v) expected subgroup size, (vi) expected size of subgroup treatment effects, and (vii) subgroup stability. We conclude with a bootstrap solution to performing post-selection inference on the selected subgroups.

Purchase Webinar Recording (11/8/19)

 

Lessons and Strategies for a Career in Academia: A Conversation

Friday, December 14, 2018
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Leslie McClure
Professor & Chair, Department of Epidemiology and Biostatistics
Dornsife School of Public Health at Drexel University

Dr. Elizabeth Stuart
Associate Dean for Education and Professor of Biostatistics, Mental Health, and Health Policy and Management
Johns Hopkins Bloomberg School of Public Health

Abstract:
As a Biostatistician, there are many paths to a successful career. Each has benefits and drawbacks and will depend on an individual's own skills and preferences. In this webinar, Drs. Elizabeth Stuart and Leslie McClure will host a conversation about their academic careers, including providing some strategies for success and describing some of the challenges they've faced. They'll consider important questions, such as: What to look for in a job? How to develop meaningful collaborations (and get out of those that are not productive)? How to prioritize activities with an eye towards promotion (e.g., collaborative and methodological projects)? How to balance teaching, research, and grant requirements? And how to balance all of that with things outside of work? However, the exact direction of the conversation will depend on the questions and engagement from webinar participants.

Purchase Webinar Recording (12/14/18)

 

Machine Learning for Health Care Policy

Friday, November 30, 2018
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Sherri Rose
Associate Professor of Health Care Policy (Biostatistics)
Harvard Medical School

Abstract:
Health care research is moving toward analytic systems that take large health databases and estimate quantities of interest both quickly and robustly, incorporating advances from statistics, machine learning, and computer science. Pressing questions in prediction and causal inference are being answered with machine learning techniques. I will give an overview of the specific challenges related to developing and deploying these statistical algorithms for health policy, including examples from the areas of health plan payment, mental health outcomes, cancer staging, and medical devices. This webinar will be accessible for graduate students with most technical derivations provided in references

Purchase Webinar Recording (11/30/18)

 

Biostatistical Methods for Wearable and Implantable Technology

Friday, October 26, 2018
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Ciprian Crainiceanu
Professor, Department of Biostatistics
Johns Hopkins University

Abstract:
Wearable and Implantable Technology (WIT) is rapidly changing the Biostatistics data analytic landscape due to their reduced bias and measurement error as well as to the sheer size and complexity of the signals. In this talk I will review some of the most used and useful sensors in Health Sciences and the ever expanding WIT analytic environment. I will describe the use of WIT sensors including accelerometers, heart monitors, glucose monitors and their combination with ecological momentary assessment (EMA). This rapidly expanding data eco-system is characterized by multivariate densely sampled time series with complex and highly non-stationary structures. I will introduce an array of scientific problems that can be answered using WIT and I will describe methods designed to analyze the WIT data from the micro- (sub-second-level) to the macro-scale (minute-, hour- or day-level) data.

Purchase Webinar Recording (10/26/18)

 

Sensitivity analysis in observational research: introducing the E-value

Friday, September 28, 2018
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Tyler VanderWeele
Professor of Epidemiology
Harvard School of Public Health

Abstract:
Sensitivity analysis is useful in assessing how robust an association is to potential unmeasured or uncontrolled confounding. This webinar introduces a new measure called the "E-value," which is related to the evidence for causality in observational studies that are potentially subject to confounding. The E-value is defined as the minimum strength of association, on the risk ratio scale, that an unmeasured confounder would need to have with both the treatment and the outcome to fully explain away a specific treatment–outcome association, conditional on the measured covariates. A large E-value implies that considerable unmeasured confounding would be needed to explain away an effect estimate. A small E-value implies little unmeasured confounding would be needed to explain away an effect estimate. The speaker and his collaborators propose that in all observational studies intended to produce evidence for causality, the E-value be reported or some other sensitivity analysis be used. They suggest calculating the E-value for both the observed association estimate (after adjustments for measured confounders) and the limit of the confidence interval closest to the null. If this were to become standard practice, the ability of the scientific community to assess evidence from observational studies would improve considerably, and ultimately, science would be strengthened.

Reference: VanderWeele, T.J. and Ding, P. (2017). Sensitivity analysis in observational research: introducing the E-value. Annals of Internal Medicine, 167:268-274.
Online E-value Calculator: https://mmathur.shinyapps.io/evalue/

Purchase Webinar Recording (9/28/2018)

 

Evidence Synthesis for Clinical trials: Use of Historical Data and Extrapolation

Friday, June 22, 2018
10:00 am – 12:00 pm Eastern

Presenters:
Sebastian Weber
Associate Director SMC
Novartis Pharma AG

Satrajit Roychoudhury
Senior Director
Pfizer Inc.

Description:
A Bayesian approach provides the formal framework to incorporate external information into the statistical analysis of a clinical trial. There is an intrinsic interest of leveraging all available information for an efficient design and analysis of clinical trials. The use of external data in trials are nowadays used in earlier phases of drug development (Trippa, Rosnerand Muller, 2012; French, Thomas and Wang, 2012; Hueber et al., 2012), occasionally in phase III trials (French et al., 2012), and also in special areas such as medical devices (FDA, 2010a), orphan indications (Dupont and Van Wilder, 2011) and extrapolation in pediatric studies (Berry, 1989). This allows trials with smaller sample size or with unequal randomization (more subjects on treatment than control). In addition, the Bayesian statistical paradigm is a natural approach for combining information across heterogeneous sources, such as different trials or the adult and pediatric data. In this webinar we'll provide a statistical framework to incorporate trial external evidence with real life examples.

During the first part of the webENAR we will introduce the meta-analytic predictive (MAP) model (Neuenschwander, 2010). The MAP model is a Bayesian hierarchical model which combines the evidence from different sources (usually studies). The MAP model provides a prediction for a future study based on available information while accounting for inherent heterogeneity in the data. This approach can be used widely in different applications of biostatistics.

In the second part of the webENAR we will focus on three key applications of the MAP approach in biostatistics, which are (i) the derivation of informative priors from historical controls, (ii) probability of success and (iii) extrapolation from adult data to pediatrics. These applications will be demonstrated using the R package RBesT, the R Bayesian evidence synthesis tools, which are freely available from CRAN. The aim of the webinar is to teach the MAP approach and enable participants to apply the approach themselves with the help of RBesT.

Purchase Webinar Recording (6/22/2018)

 

Programming with hierarchical statistical models: An introduction to the BUGS-compatible NIMBLE system for MCMC and more

Friday, April 13, 2018
11:00 am – 1:00 pm Eastern

Presenter:
Chris Paciorek
Adjunct Professor, Statistical Computing Consultant
Department of Statistics
University of California, Berkeley

Description:
This webinar will introduce attendees to the NIMBLE system for programming with hierarchical models in R. NIMBLE
(r-nimble.org) is a system for flexible programming and dissemination of algorithms that builds on the BUGS language for declaring hierarchical models. NIMBLE provides analysts with a flexible system for using MCMC, sequential Monte Carlo and other techniques on user-specified models. It provides developers and methodologists with the ability to write algorithms in an R-like syntax that can be easily disseminated to users. C++ versions of models and algorithms are created for speed, but these are manipulated from R without any need for analysts or algorithm developers to program in C++.

While analysts can use NIMBLE as a drop-in replacement for WinBUGS or JAGS, NIMBLE provides greatly enhanced functionality in a number of ways. The webinar will first show how to specify a hierarchical statistical model using BUGS syntax (including user-defined function and distributions) and fit that model using MCMC (including user customization for better performance). We will demonstrate the use of NIMBLE for biostatistical methods such as semiparametric random effects models and clustering models. We will close with a discussion of how to use the system to write algorithms for use with hierarchical models, including building and disseminating your own methods.

Purchase Webinar Recording (4/13/2018)

 

Incorporating Patient Preferences into Regulatory Decision Making

Friday, December 1, 2017
11:00 am - 1:00 pm Eastern

Presenter:
Telba Irony, PhD
Center for Biologics Evaluation and Research
FDA

Description:
Regulatory authorities and patient advocacy groups have been paving the way towards engaging patients in medical product development and regulatory review. These efforts gave rise and relevance to the development of the Science of Patient Input, or SPI. SPI consists of scientifically valid qualitative and quantitative methods for capturing patient perspective information to incorporate it into product development and regulatory decision making. Two types of patient input, Patient Reported Outcomes (PRO) and Patient Preference Information (PPI) are expected to be captured in accordance with applicable scientific and statistical standards and best practices, and statisticians have a large role to play.

A PRO is a measurement based on a report of a patient’s health status that comes directly from the patient, without interpretation of the patient’s report by a clinician or anyone else. Some symptoms or other unobservable concepts known only to the patient, such as pain or fatigue, can only be measured by PRO measures.

PPI is a patient’s expression of desirability or value of one course of action or selection in contrast to others. It focuses on assessing the importance, or preferences, that patients place on the benefits, harms and other aspects of treatments.

In this Webinar we will introduce key elements concerning elicitation and use of patient preferences (PPI) to inform regulatory decision making. As an example, we will present a study commissioned by the FDA to elicit obese patients’ preferences in choosing weight-loss devices and show how these preferences can be used to inform regulatory decision making. We will describe the weight-loss device survey and present the survey results, which have been used to develop a decision-aid tool for regulatory reviewers. The tool provides estimates of patients’ benefit-risk tradeoff preferences and also stratifies patients according to their risk-tolerance. We will conclude the Webinar by sharing experiences in using patient preferences in the regulatory process and talking about best statistical practices for eliciting and using patient preference information.

Purchase Webinar Recording (12/1/17)

 

Basket Trials for Development of Targeted Treatment Regimens

Friday, November 17, 2017
11:00 am – 1:00 pm Eastern

Presenter:
Dr. Mithat Gönen
Chief, Biostatistics Service
Memorial Sloan Kettering Cancer Center

Description:
Cancer clinical trials have traditionally been designed specific to a disease site (breast, lung, colon etc). This paradigm is being challenged by the advent of targeted treatments, regimens targeting molecular alterations in cancer cells. Since targeted treatments are not site-specific the trials evaluating them increasingly include multiple sites where the target is expressed. These trials are often called basket trials. In this WebENAR we will present several possible designs for basket trials: parallel design, aggregation design and hierarchical model-based design; comparing their operating characteristics, strengths and weaknesses. Although their applications have mostly been in oncology so far, basket trials can be used in any disease where targeted treatments can be used in molecularly defined subgroups. We will give examples of publicly available software that can be used to design and analyze basket trials.

Purchase Webinar Recording (11/20/17)

 

Multi-state Models: Methods and Software

Friday, October 20, 2017
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Christopher Jackson
Senior Statistician, MRC Biostatistics Unit, University of Cambridge
School of Clinical Medicine, Cambridge Institute of Public Health

Description:
Multi-state models are stochastic processes which describe how an individual moves between a set of discrete states in continuous time. They have been used for two broad classes of data. Firstly, for "panel data": intermittent observations of the state at a finite series of times, for a set of individuals, where transition times are not known. Secondly, for times to multiple events for a set of individuals, so that the state at any time is known. Combinations or slight variants of these two data types are also possible.

Purchase Webinar Recording (10/20/17)

 

Data-driven Decision Making with Application to Precision Medicine

Friday, June 30, 2017
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Eric Laber
Associate Professor
North Carolina State University

Description:
In this webinar we will cover the methodological and computational underpinnings of data-driven decision making with application to precision medicine. Planned topics covered include: (i) formalizing optimal decision making through potential outcomes; (ii) regression and classification-based methods for single-stage decision problems; (iii) approximate dynamic programming and direct-search methods for multi-stage decision problems; (iv) interpretability and the research-practice gap; and (v) current research topics and open problems. We do not assume that participants have any prior exposure to these topics; however, it is assumed that participants have at least a masters-level understanding of biostatistics.

Purchase Webinar Recording (6/30/17)

 

The Overlap between Statisticians and Pharmacometricians in Clinical Drug Development and a Case Study

Friday, May 19, 2017
10:00 am – 12:00 pm Eastern

Presenter:
Kenneth G. Kowalski, MS
Kowalski PMetrics Consulting, LLC

Wenping Wang, PhD
Novartis Pharmaceuticals Corporation

Description:
This WebENAR will be presented in two parts. The first part will focus on a commentary presented by Ken Kowalski discussing the overlap between statisticians and pharmacometricians working in clinical drug development. Individuals with training in various academic disciplines including pharmacokinetics, pharmacology, engineering and statistics, to name a few, have pursued careers as pharmacometricians. While pharmcometrics has benefitted greatly from advances in statistical methodology, there is considerable tension and skepticism between biostatisticians and pharmacometricians as they apply their expertise to drug development applications. This talk explores some of the root causes for this tension and provides some suggestions for improving collaborations between statisticians and pharmcometricians. The talk concludes with a plea for more statisticians to consider careers as pharmacometrics practitioners. The second part of the WebENAR will highlight a case study presented by Wenping Wang illustrating the application of pharmacokinetic-pharmacodynamic modeling of the time to first flare to support dose justification of Canakinumab in a sBLA submission. The case study will conclude with some observations regarding team interactions between statisticians and pharmacometricians that resulted in a successful sBLA submission.

Purchase Webinar Recording (5/19/17)

 

Evaluation and Use of Surrogate Markers

Friday, April 21, 2017
11:00 am – 1:00 pm Eastern

Presenter:
Dr. Layla Parast
Statistician
RAND Corporation

Description:
The use of surrogate markers to estimate and test for a treatment effect has been an area of popular research. Given the long follow-up periods that are often required for treatment or intervention studies, appropriate use of surrogate marker information has the potential to decrease required follow-up time. However, previous studies have shown that using inadequate markers or making inappropriate assumptions about the relationship between the primary outcome and the surrogate marker can lead to inaccurate conclusions regarding the treatment effect. Many of the available methods for identifying, validating and using surrogate markers to test for a treatment effect tend to rely on restrictive model assumptions and/or focus on uncensored outcomes. In this course, I will describe different approaches to quantify the proportion of treatment effect explained by surrogate marker information in both a non-survival outcome setting and censored survival outcome setting. One described approach will be a nonparametric method that can accommodate a setting where individuals may experience the primary outcome before the surrogate marker is measured. I will illustrate the procedures using an R package available on CRAN to examine potential surrogate markers for diabetes with data from the Diabetes Prevention Program.

Purchase Webinar Recording (4/21/17)

 

How Credible are Your Conclusions about Treatment Effect When There are Missing Data? Sensitivity Analyses for Time-to-event and Recurrent-event Outcomes

Friday, February 24, 2017
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Michael O'Kelly
Dr. Bohdana Ratitch
Dr. Ilya Lipkovich
Center for Statistics in Drug Development
Quintiles

Description:
Most experiments have missing data. When there are missing data, it is useful to provide sensitivity analyses to allow the reader of the account of the research to assess the robustness to the missing data of any conclusions made. Using the pattern-mixture framework, a variety of assumptions can be implemented with regard to categories of missing outcomes. Assumptions that would tend to undermine the alternative hypothesis can be especially useful for assessing robustness of conclusions. Multiple imputation (MI) is one quite straightforward way of implementing such pattern-mixture approaches. While MI is a standard tool for continuous outcomes, recently researchers have come up with ways of implementing MI for other outcomes, such as time-to-event and recurrent-event outcomes. This webinar describes a number of these new applications of the MI idea. The strengths and weaknesses of these approaches are described and illustrated via examples and simulations.

This webinar was not recorded and is not available for on-demand purchase.

 

Introduction to Clinical Trial Optimization

Friday, February 3, 2017
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Alex Dmitrienko
Founder & President
Mediana Inc.

Description:
This webinar focuses on a broad class of statistical problems related to optimizing the design and analysis of Phase II and III trials (Dmitrienko and Pulkstenis, 2017). This general topic has attracted much attention across the clinical trial community due to increasing pressure to reduce implementation costs and shorten timelines in individual trials and development programs.

The Clinical Scenario Evaluation (CSE) framework (Benda et al., 2010) will be described to formulate a general approach to clinical trial optimization and decision-making. Using the CSE approach, main objectives of clinical trial optimization will be formulated, including selection of clinically relevant optimization criteria, identification of sets of optimal and nearly optimal values of the parameters of interest, and sensitivity assessments. Key principles of clinical trial optimization will be illustrated using a problem of identifying efficient and robust multiplicity adjustment strategies in late-stage trials (Dmitrienko et al., 2009; Dmitrienko, D’Agostino and Huque, 2013; Dmitrienko, Paux and Brechenmacher, 2015).

Software tools for applying optimization methods will be presented, including R software (Mediana package) and Windows applications with a graphical user interface (MedianaFixedDesign application).

Purchase Webinar Recording (2/3/17)

 

Design and Analysis of Genome-Wide Association Studies

Friday, December 9, 2016
10:00 am – 12:00 pm Eastern

Presenter:
Dr. Nilanjan Chatterjee
Bloomberg Distinguished Professor
Department of Biostatistics, Bloomberg School of Public Health
Department of Oncology, School of Medicine
Johns Hopkins University

Description:
Decreasing cost of large scale genotyping and sequencing technologies is fuelling investigation of association between complex traits and genetic variants across the whole genome using studies of massive sample sizes. Recent genome-wide association studies (GWAS) focused on common variants have already led to the discoveries of thousands of genetic loci across variety of complex traits, including chronic diseases such as cancers, heart diseases and type-2 diabetes. Future studies of less common and rare variants hold further promise for discovery of new genetic loci and better understanding of causal mechanisms underlying existing loci. The webinar will provide brief review of some state of the art design and analysis issues faced in the field. The topics will include sample size requirement and power calculations, methods for single- and multi-marker association testing, estimation of heritability and effect-size distribution, techniques for pleiotropic and Mendelian randomization analyses and genetic risk prediction.

Purchase Webinar Recording (12/9/16)

 

Nonparametric Bayes Biostatistics

Friday, October 28, 2016
10:00 am – 12:00 pm Eastern

Presenter:
Dr. David Dunson
Arts & Sciences Professor of Statistical Science, Mathematics and Electrical & Computer Engineering
Duke University

Description:
This webinar will provide an introduction to the practical use of nonparametric Bayesian methods in the analysis and interpretation of data from biomedical studies. I will start with a very brief review of the Bayesian paradigm, rapidly leading into what is meant by "Nonparametric Bayes." I'll then describe some canonical nonparametric Bayes models, including Dirichlet process mixtures and Gaussian processes. Basic practical properties and approaches for computation will be sketched, and I'll provide a practical motivation through some biomedical applications ranging from genomics to epidemiology to neuroscience. I'll finish up by describing some possibilities in terms of more advanced models that allow the density of a response variable to change flexibly with predictors, while providing practical motivation and implementation details.

 

Pragmatic Trials in Public Health and Medicine

Friday, May 20, 2016
11:00 am- 1:00 pm Eastern

Presenters:
David M. Murray, Ph.D.
Associate Director for Prevention
Director, Office of Disease Prevention
Office of the Director
National Institutes of Health

Description:
This webinar will review key issues and their solutions for pragmatic trials in public health and medicine. Pragmatic trials are used increasingly in health care settings to help clinicians choose between options for care. They often involve group- or cluster-randomization, though alternatives to randomized trials are also available. Many current trials rely upon electronic health records as the major source for data. These studies face a variety of challenges in the development and delivery of their interventions, research design, informed consent, data collection, and data analysis. This webinar will review these issues both generally and using examples from the Health Care Systems Collaboratory. The HCS Collaboratory is an NIH funded consortium of nine pragmatic trials that address a variety of health issues and outcomes, all conducted within health care systems, all relying on electronic health records as their primary source of data, with most implemented as a group- or cluster-randomized trial.

 

Analytic Methods for Functional Neuroimaging Data

Friday, April 15, 2016
10:00 am- 12:00 pm Eastern

Presenters:
F. DuBois Bowman
Dr. Daniel Drake
Dr. Ben Cassidy
Department of Biostatistics, Mailman School of Public Health
Columbia University

Description:
Brain imaging scanners collect detailed information on brain function and various aspects of brain structure. When used as a research tool, imaging enables studies to investigate brain function related to emotion, cognition, language, memory, and responses to numerous other external stimuli, as well as resting-state brain function. Brain imaging studies also attempt to determine the functional or structural basis for psychiatric or neurological disorders and to examine the responses of these disorders to treatment. Neuroimaging data, particularly functional images, are massive and exhibit complex patterns of temporal and spatial dependence, which pose analytic challenges. There is a critical need for statisticians to establish rigorous methods to extract information and to quantify evidence for formal inferences. In this webinar, I briefly provide background on various types of neuroimaging data (with an emphasis on functional data) and analysis objectives that are commonly targeted in the field. I also present a survey of existing methods aimed at these objectives and identify particular areas offering opportunities for future statistical contribution.

 

Bayesian Population Projections

Friday, February 12, 2016
10:00 am - 12:00 pm Eastern

Presenter:
Adrian E. Raftery
Professor of Statistics and Sociology
University of Washington

Description:
Projections of countries' future populations, broken down by age and sex, are widely used for planning and research. They are mostly done deterministically, but there is a widespread need for probabilistic projections. I will describe a Bayesian statistical method for probabilistic population projections for all countries. These new methods have been used by the United Nations to produce their most recent population projections for all countries.

 

Regulatory Perspective on Subgroup Analysis in Clinical Trials

December 4, 2015
10:00 am - 12:00 pm Eastern

Presenters:
Dr. Mohamed Alosh & Dr. Kathleen Fritsch
Division of Biometrics III, Office of Biostatistics, OTS, CDER, FDA

Description:
For a confirmatory clinical trial that established treatment efficacy in the overall population, subgroup analysis aims to investigate the extent of benefits from the therapy for the major subgroups. Consequently, findings from the subgroup analysis play a major role in interpreting the trial results. This presentation focuses on two areas related to subgroup analysis in a confirmatory clinical trial: (i) investigating consistency of treatment effect across subgroups, and (ii) designing a clinical trial with the objective of establishing treatment efficacy in a targeted subgroup in addition to the overall population. The presentation also outlines the regulatory guidelines for subgroup analysis in such trials and provides examples of clinical trials where subgroup analysis played a role in determining the population for treatment use.

 

Reproducible Research: The Time is Now

Friday, November 20, 2015
10:00 am - 12:00 pm Eastern

Presenter:
Dr. Keith Baggerly
The University of Texas MD Anderson Cancer Center

Description:
The buzz phrase "Reproducible Research" refers to studies where the raw data and code supplied are enough to let a new investigator exactly match the reported results without a huge amount of effort. "Replicable Research" refers to studies whose methods, when applied to new data, give rise to qualitatively similar results. Particularly as experiments get bigger, more involved, and more expensive, reproducibility should precede replication. Unfortunately, more attention is now being focused on such issues due to some high-profile failures.

In this talk, we first illustrate the issues with some case studies from oncology showing the types of things that can go wrong, the simple nature of the most common mistakes, and what the implications can be: e.g., treating patients incorrectly. We then give some point estimates of how widespread the problems of reproducibility and replicability are thought to be, and discuss some additional problems associated with the replication. We survey tools introduced in the past few years which have made assembling reproducible studies markedly easier, discuss considerations to be applied when considering replication, and give pointers to some resources for further information.

 

The Statistical Analysis of fMRI Data

Friday, June 26, 2015
10:00 am to 12:00 pm Eastern

Presenter:
Martin Lindquist
Professor
Department of Biostatistics
Johns Hopkins University

Description:
Functional Magnetic Resonance Imaging (fMRI) is a non-invasive technique for studying brain activity. During the past two decades fMRI has provided researchers with an unprecedented access to the inner workings of the brain, leading to countless new insights into how the brain processes information. The field that has grown around the acquisition and analysis of fMRI data has experienced a rapid growth in the past several years and found applications in a wide variety of areas. This webinar introduces fMRI and discusses key statistical aspects involved in the analysis of fMRI data. Topics include: (a) an overview of the acquisition and reconstruction of fMRI data; (b) overview of the physiological basis of the fMRI signal; (c) common experimental designs; (d) pre-processing steps; (d) methods for localizing areas activated by a task; (e.) connectivity analysis; and (f.) prediction and brain decoding.

 

Sparse Logical Models for Interpretable Machine Learning

Friday, May 8, 2015
10:00 am to 12:00 pm Eastern

Presenter:
Cynthia Rudin, PhD, Associate Professor of Statistics, MIT CSAIL and Sloan School of Management, Massachusetts Institute of Technology

Description:
Possibly *the* most important obstacle in the deployment of predictive models is the fact that humans simply do not trust them. If it is known exactly which variables were important for the prediction and how they were combined, this information can be very powerful in helping to convince people to believe (or not believe) the prediction and make the right decision. In this talk I will discuss algorithms for making these non-black box predictions including:

1. "Bayesian Rule Lists" - This algorithm builds a decision list using a probabilistic model over permutations of IF-THEN rules. It competes with the CART algorithm for building accurate-yet-interpretable logical models. It is not a greedy algorithm like CART.


2. "Falling Rule Lists" - These are decision lists where the probabilities decrease monotonically along the list. These are really useful for medical applications because they stratify patients into risk categories from highest to lowest risk.


3. "Bayesian Or's of And's" - These are disjunctions of conjunction models (disjunctive normal forms). These models are natural for modeling customer preferences in marketing.


4. "The Bayesian Case Model" - This is a case-based reasoning clustering method. It provides a prototypical exemplar from each cluster along with the subspace that is important for the cluster.

 

Statistical Issues in Comparative Effectiveness Research

Friday, February 20, 2015
11:00 am to 1:00 pm EST

Presenter:
Sharon-Lise Normand, Department of Health Care Policy, Harvard Medical School & Department of Biostatistics, Harvard School of Public Health

Description:
Comparative Effectiveness Research (CER) refers to a body of research that generates and synthesizes evidence on the comparative benefits and harms of alternative interventions to prevent, diagnose, treat, and monitor clinical conditions, or to improve the delivery of health care. The evidence from CER is intended to support clinical and policy decision making at both the individual and the population level. While the growth of massive health care data sources has given rise to new opportunities for CER, several statistical challenges have also emerged. This tutorial will provide an overview of the types of research questions addressed by CER, review the main statistical methodology currently utilized, and highlight areas where new methodology is required. Inferential issues in the "big data" context are identified. Examples from cardiology and mental illness will illustrate methodological issues.

 

Statistical Challenges in Genomics High Throughput Data

Friday, January 30, 2015
11:00 am to 1:00 pm EST

Presenter: Rafa Irizarry, PhD
Professor of Biostatistics and Computational Biology at the Dana Farber Cancer Center
Professor of Biostatistics at the Harvard School of Public Health
http://rafalab.dfci.harvard.edu/

Description:
In this webinar I will give an overview of genomics technologies and the challenges arising when analyzing the data they produce. Specifically, I will focus on microarrays and next generation sequencing technologies. We will cover statistical issues related to preprocessing, normalization, detecting differential expression, and dealing with batch effects.

 

An Introduction to Dynamic Treatment Regimes

December 5, 2014
11:00 am to 1:00 pm (EST)

Presenter: Marie Davidian, PhD North Carolina State University

Description: Treatment of patients with chronic diseases or disorders in clinical practice involves a series of decisions made over time. Clinicians adjust, change, modify, or discontinue therapies based on the patient's observed progress, side effects, compliance, and so on, with the goal of "personalizing" treatment to the patient in order to provide the best care. The decisions are typically based on synthesis of the available information on the patient using clinical experience and judgment.

A "dynamic treatment regime," also referred to as an "adaptive treatment strategy," is a set of sequential rules that dictate how to make decisions on treatment of a patient over time. Each rule corresponds to a key decision point at which a decision on which treatment action to take from among the available options must be made. Based on patient information, the rule outputs the next treatment action. Thus, a dynamic treatment regime is an algorithm that formalizes the way clinicians manage patients in practice.

In this presentation, we introduce the notion of a dynamic treatment regime and an appropriate statistical framework in which treatment regimes can be studied. We demonstrate how statistical inference may be made on the effects of different regimes based on data from so-called sequential, multiple assignment, randomized trials (SMARTs). We conclude with a discussion of current challenges, including the development of "optimal" treatment regimes. The material presented is ideal background for the shortcourse on personalized medicine and optimal dynamic treatment regimes to be offered at the ENAR Spring Meeting in March 2015.