All Times ET
Viewing session type: Tutorial
Back to search menu
Thursday, February 3
Thu, Feb 3
9:00 AM - 11:00 AM
Virtual
T01 - Network Analysis to Solve Business Problems
Tutorial
Instructor(s): Carlos Pinheiro, SAS Institute
Network analysis includes graph theory algorithms that can augment data mining and machine learning. In many practical applications, pairwise interaction between the entities of interest in the model often plays an important role. Network analysis goes beyond traditional clustering and predictive models to identify patterns in business data, including entities’ behavior based on their relationships. Network analysis can be employed to avoid churn, diffuse products and services, detect fraud and abuse, identify anomalies, and many other applications, in a wide range of industries such as communications and media, banking, insurance, retail, utilities, and travel and transportation.
Outline & Objectives
Section 1: Fundamental Concepts in Network Analysis
Introduction
Concepts about network analysis
The type of data for network building and network analysis
Section 2: Sub-Networks
Connected components
Bi-connected components
Community detection
Reach
Core
Section 3: Centrality Measures
Degree
Influence
Clustering coefficient
Closeness
Betweenness
Hub
Authority
Eigenvector
PageRank
About the Instructor
Carlos Pinheiro is a Principal Data Scientist at SAS, US and a Visiting Professor at Data ScienceTech Institute, France. He led analytical teams at Embratel, Brasil Telecom and Oi, worked as a Senior Data Scientist for EMC on network analytics, optimization, and text analytics projects, and as a Lead Data Scientist for Teradata on machine learning projects. Dr. Pinheiro has a BSc in Mathematics and Computer Science, a MSc in Computing, and a DSc in Engineering from Federal University of Rio de Janeiro. He has completed a series of postdoctoral research terms in different fields, including Dynamic Systems at IMPA, Brazil, Social Network Analysis at Dublin City University, Ireland, Transportation Systems at Université de Savoie, France, Dynamic Social Networks and Human Mobility at KU Leuven, Belgium, and Urban Mobility and Multi-modal Traffic at Fundação Getúlio Vargas, Brazil. He is author of Social Network Analysis in Telecommunications and Heuristics in Analytics: A Practical Perspective of What Influence Our Analytical World, both published by John Wiley Sons, Inc, and Introduction to Statistical and Machine Learning Methods for Data Science, SAS Press.
Relevance to Conference Goals
This tutorial provides a practical perspective about how to use network analysis to solve real-world problems, including hands-on demonstrations on the algorithms and case studies. The course focuses on the data science approach to solve business problems, combining different techniques to evaluate the problem and propose optimal solutions.
Thu, Feb 3
9:00 AM - 11:00 AM
Virtual
T02 - Fundamentals of Study Design and Analysis Plans for Biomarker Research
Tutorial
Instructor(s): Douglas Landsittel, Indiana University, Bloomington
Biomarker studies are ubiquitous and critically significant across almost every discipline and stage of research. Biomarkers are often necessary for diagnosis of disease status, prognosis for future outcomes, and prediction of treatment response. They are also critical for understanding mechanisms of disease, monitoring disease progression, and identifying high risk populations most likely to benefit from interventions and medical treatments. In addition, surrogate biomarkers are often necessary when the true clinical endpoint is either impractical to measure or develops too slowly for effective intervention. However, despite the critical role of biomarker studies, many key aspects of their study designs and statistical analysis plans (SAPs) are poorly understood, thus leading to suboptimal funding proposals and poorly designed study protocols. This workshop describes the necessary concepts and key steps in study design and SAPs for biomarker research. Those concepts and specific steps are illustrated through description of challenges and approaches for two ongoing multi-site studies in polycystic kidney disease and severe acute respiratory infections (including COVID-19).
Outline & Objectives
The objectives of this half-day workshop are to describe, and illustrate examples of, best practices in biomarker study design and statistical analysis planning. Participants will gain skills to effectively design and write analysis plans for proposals and study protocols.
The content assumes only a basic knowledge of regression and study design (e.g. introductory biostatistics or epidemiology course).
Workshop Topics:
Part I: Introduction:
1) Definitions and applications
2) Biomarker panels, signatures and other high dimensional data
3) Case studies in polycystic kidney disease and COVID-19
Part II: Overview of Regression for Biomarker Analysis:
1) Goals of regression
2) Models for classification and prediction
3) Evaluation of accuracy
Part III: Types of Biomarkers and Associated Statistics:
1) Why it matters
2) Differential expression, correlation, diagnosis, prognosis, and response prediction
3) Surrogate markers and clinical endpoints
Part IV: Study Designs and Phases of Biomarker Research
1) Subject selection, timing of measurements, and randomization
2) Classifying phases of biomarker development
3) The need for multiple studies.
Conclusions
About the Instructor
Dr. Landsittel is the Professor and Chair of Epidemiology and Biostatistics in the School of Public Health at Indiana University Bloomington. He has published nearly 150 peer-reviewed papers, many of which are focused on biomarker studies. Previously, he has served as the Associate Director of the Biostatistics Facility for the Hillman Cancer Center, Associate Director of the Center for Research on Healthcare Data Center, and Director of Biostatistics (for Research) for the Starzl Transplant Institute. He has also been appointed to study sections, and is the Chair of the Safety and Occupational Health Study Section, and has served on numerous other biomarker-related expert panels.
Relevance to Conference Goals
The Conference on Statistical Practice seeks to engage “statistical practitioners and data scientists” in real-world problems, including in the area of study design. This proposal addresses study design challenges in one of the most critical and common areas across nearly disciplines: biomarker research.