Title : Surviving the Big Data challenge - Why Data Science needs Statisticians for making AI Effective

Course Facilitator : Prof. Arnab Kumar Laha, Indian Institute of Management Ahmedabad

Prof. Arnab Kumar Laha

Abstract :

From the onset of the 21st century we have seen a rapid growth of efficiencies of technologies for collection and storage of data. This has resulted in huge volumes of data of different kinds being acquired by commercial and scientific organisations across the world. Many organisations are now gearing up to use the gathered data to build useful Decision Support Systems (DSS) popularly called AI systems. Many of these DSS are sought to be given capabilities that supplant or even eliminate the need for a human interface. This is a tough task in many situations given the uncertainties associated with the acquired data. Moreover, the nature of the acquired data is often complex involving multiple data types that make the inference problem a substantial technical challenge. Some of these DSS need to provide results in real-time making human supervision of output nearly impossible. Moreover, since the data gathering process is often automated, the data scientist needs a helping hand from statisticians to develop procedures that are reliable and robust to the presence of outliers. In some business situations involving real-time decision making the distribution of the incoming data may change quite abruptly over time and it is expected that the implemented procedures are such that the quality of the output of the DSS is not affected by such changes.

A rather vast amount of Big Data has been and is being collected for purposes other than for which it is being used. This makes the analysis of such data fraught with difficulties. The non-random nature of the data raises fundamental questions regarding generalisability of the results obtained from the analysis of such data. However, the business value proposition of using Big Data Analytics (BDA) is maximum when the results obtained can be generalised to the target population. Statisticians have a significant role to play in devising methods that make generalisable inference from such data a possibility. What makes a statistician's role even more important is the fact that one may need to devise innovative procedures unique to a particular dataset to attain a semblance of generalisability.

Uncertainty quantification for inference on Big Data is a huge area of research where statisticians are uniquely capable of contributing, The need for uncertainty quantification cannot be overstated for AI systems which runs without human supervision but provides actionable outputs for use by humans. Without proper uncertainty quantification AI systems may produce outputs which may mislead humans or may cause social outrage as had happened in the recent past. Through proper probabilistic modelling of Big Data statisticians can help in quantifying the uncertainty.

In this workshop, we discuss the above issues with real-life examples drawn from business, scientific and public policy domains.

Pre-requisites and Target Audience :

Statisticians and Data Scientists who are involved in (or has plans to be involved in) analysis of Big Data and development of AI based DSS.

Title : Statistical Methods for Geospatial Analysis

Course Facilitator : Dr. Pradeep Mohan, SAS Institute

Dr. Pradeep Mohan

Dr. Pradeep Mohan (Pradeep.Mohan@sas.com) is a Senior Research Statistician Developer in the Advanced Analytics Division at SAS Institute Inc. Pradeep’s main responsibility is developing and enhancing the geospatial analysis capabilities of SAS® software with a focus on geostatistical modeling (in the VARIOGRAM, KRIGE2D, and SIM2D procedures), spatial point process modeling (in the SPP procedure), and other general-purpose geospatial computing utilities. Before joining SAS in 2012, Pradeep received his Master’s and PhD Degrees in Computer Science with a strong emphasis on the computing aspects of geospatial analysis from the University of Minnesota, Twin Cities.

Abstract :

This half-day course introduces different examples of geospatial phenomena, their observational data (raster and vector), and their statistical interpretation as geospatial processes (the geostatistical process, the point process, and the lattice process). The course then delves deeply into different aspects of geospatial process modeling, including data exploration, diagnostics for spatial dependency or spatial heterogeneity (or both), modeling techniques to account for both spatial dependency and spatial heterogeneity in different forms, model criticism and validation using residuals, and, finally, model selection. Computational aspects of handling geospatial data for different geospatial process modeling techniques will be highlighted throughout the course. Although the course will touch on all three types of geospatial process modeling, the primary focus will be on the geostatistical process and the spatial point process. We will use SAS® software to illustrate different concepts during the course.

Prerequisites and target audience : :

A basic understanding of geospatial phenomena and geospatial observational data will be helpful but not necessary. A basic understanding of statistical distributions, linear regression, and generalized linear models is assumed. This course is intended for a broad audience who are interested in statistics, computer science, and the geospatial sciences. Specific groups include (but are not limited to) the following:

Software Packages :

We will use SAS University Edition for different examples. SAS University Edition can be downloaded for free from the following link:

Where necessary, SAS macros will also be used for instruction. These macros will be provided to participants in advance.

Reference Text :

The following textbooks are useful as references for this course:

Title : Moving your career forward – Establishing Professional Goals

Course Facilitator : Dr. Mahesh Iyer, Sineflex Solutions LLP

Dr. Mahesh Iyer

Mahesh Iyer is a co-founder of Sineflex Solutions LLP, a consulting company focused on accelerating innovation in the healthcare space. Mahesh has a Ph.D. in Statistics from Temple University, Philadelphia and has over 20 years of experience as a statistician in various Pharma companies including Boehringer Ingelheim, Bristol Myers Squibb and Novartis. Over the course of his career, he has held various positions starting of as a trial statistician, moving on to lead projects, and eventually build teams and act as the Global Head for Stats and Analytics teams for Novartis. He is also passionate about teaching and mentoring; he is adjunct faculty at multiple universities, and is an EQi2.0 certified coach. He will be facilitating this workshop.

Abstract :

Many of us get to a point in our careers, where we start wondering, what next? There is confusion about how to take our careers forward; should we focus on enhancing our statistical skills, or our soft skills? Should we look to develop as a statistician, or a manager, or both?

Professional goals are crucial elements in career planning and achievement. Like a compass, concrete goals help us get back on our chosen trajectory when the unavoidable distractions from inside and outside work push us off. Goals can help us chart our path and even help with leveraging experiences as growth opportunities that might otherwise seem more like “drifting” than purposeful, forward momentum. This course will provide instruction and guidance on setting, documenting, and evaluating progress toward reasonable short-, medium-, and long-term professional goals. During the course, we will discuss identification and formulation of goals, sequencing; the importance and generation of evidence of achievement, and use of this evidence to create a plan for next-stage goal setting. This workshop will also explore the soft skills required for achieving our set goals, and how to go about acquiring them. This interactive workshop will comprise a few brief lectures, individual work, and small- and large-group discussions.

Pre-conference work will include submission of at least one specific short-term (up to 1 year to complete), medium-term (1–3 years to complete), and long-term (3–10 years to complete) goal. These professional goals will not be shared unless participants choose to, but they are essential to focus workshop activities that are relevant for each individual. Participants should bring a laptop and current version of their CV or résumé, as well as come prepared for active and reflective participation.

Pre-requisites and Target Audience :

Statisticians and data scientists who have had at least 3 years of work experience, and are actively looking to chart out their career journey

Title : Workshop on Quantile Regression

Course Facilitator : Professor Roger Koenker, University of Illinois at Urbana Champaign and University College London
Professor Lan Wang, University of Minnesota
Assistant Professor Naveen N. Narisetty, University of Illinois at Urbana-Champaign

Professor Roger Koenker

Professor Roger Koenker is part of the faculty of Economics and Statistics at the University of Illinois at Urbana Champaign and at University College London. Prof. Koenker is the inventor of quantile regression along with his co-author Prof. Gib Bassett. He received many esteemed honors including the Emanuel and Carol Parzen Prize in 2010 for Statistical Innovation for his contribution to the field and for "pioneering and expositing quantile regression." He published two seminal books, Quantile Regression (2005), and Handbook of Quantile Regression (2017) along with a number of highly influential papers on quantile regression and empirical Bayes. Prof. Koenker maintains the popular R package quantreg that is used by a majority of practitioners both in industry and academia.

Professor Lan Wang

Professor Lan Wang is a faculty member in the School of Statistics at the University of Minnesota, Twin Cities. She is one of the foremost leaders on quantile regression and published a number of highly cited papers in all the top journals in Statistics. Prof. Wang was named the Fellow of the American Statistical Association in 2018 and as Fellow of the Institute of Mathematical Statistics in 2017. She has been an Associate Editor for JASA, Annals of Statistics, JRSSB, and Biometrics.

Assistant Professor Naveen Naidu Narisetty

Assistant Professor Naveen Naidu Narisetty is a faculty in the Department of Statistics at the University of Illinois at Urbana-Champaign (UIUC). Prof. Narisetty received his Bachelors and Masters from the Indian Statistical Institute, Kolkata and PhD from University of Michigan, Ann Arbor. He has a broad research interests including Bayesian methodology and computation, high dimensional statistics, and quantile regression. Prof. Narisetty received several honors including Fellowship from the Center for Advanced Study at UIUC and ProQuest Distinguished Dissertation Award from Michigan.

Abstract :

Quantile regression models (Koenker and Bassett, 1978; Koenker, 2005; Koenker, 2017) provide a flexible and powerful alternatives to mean regression models and have found widespread use in various application areas including biology, economics, finance, marketing, and environmental sciences. Quantile regression models offer a natural and intuitive way to model and interpret heterogeneity in the data and have become an important toolbox for both theoretical and applied researchers.

While standard regression methods are regularly taught in undergraduate and graduate courses, quantile regression is rarely taught as part of standard curriculum of statistics undergraduate and graduate programs. This workshop aims to make the powerful techniques of quantile regression available to a broad range of researchers and professionals by introducing the background, fundamental principles, methods, computational tools, software implementation along with concrete illustrations to specific applications. An attendee of the workshop is expected to gain both a broad overview of the important statistical technology quantile regression has to offer as well as some deep insights and intuition about how quantile regression methods work. Topics covered will include basics of Quantile regression (What, Why, and How), Statistical Inference for Quantile regression, Computation and examples, Quantile regression for censored data, Nonparametric Quantile regression, and Bayesian Quantile Regression. These topics will be demonstrated with examples and applications.

Pre-requisites and Target Audience :

Familiarity with the basics of statistics inference and regression techniques. The workshop is expected to be beneficial to experienced practitioners of statistics, graduate students, and academic researchers.

Title : R Conference At IISA

Invited Speakers : Susan Holmes, Professor of Statistics, Stanford University.
Martin Morgan, Research Professor, Biostatistics, SUNY, Buffalo and Director of the Bioconductorproject.

Organizers : Balasubramanian Narasimhan, (Chair) is a Senior Research Scientist in the Department of Biomedical Data Sciences and in the Department of Statistics at Stanford University.
Megha Patnaik, Visiting Assistant Professor at the Economics and Planning Unit of the Indian Statistical Institute, New Delhi
Deepayan Sarkar, Associate Professor in the Theoretical Statistics and Mathematics Unit of the Indian Statistical Institute, New Delhi.
Susan Thomas, Assistant Professor at the Indira Gandhi Institute of Development Research, Mumbai.

Balasubramanian Narasimhan

Balasubramanian Narasimhan is a Senior Research Scientist in the Department of Biomedical Data Sciences and in the Department of Statistics at Stanford University. He is also the Director of the Data Coordinating Center in the Stanford School of Medicine. He received his PhD from the Department of Statistics, Florida State University in 1991. He is a member of the R Foundation and previously organized the useR! 2016 conference. His research interests are in statistical computing, machine learning, adaptive clinical trial design and bioinformatics.

Megha Patnaik

Megha Patnaik is visiting Assistant Professor at the Economics and Planning Unit of the Indian Statistical Institute, New Delhi. She received her doctorate in Economics from Stanford University in June 2017. Her research interests are Labour, Finance, Macroeconomics and Entrepreneurship. She completed her BSc. in Mathematics from St. Stephen’s College and MSc. in Econometrics and Mathematical Economics from the London School of Economics.

Deepayan Sarkar

Deepayan Sarkar is an Associate Professor in the Theoretical Statistics and Mathematics Unit of the Indian Statistical Institute, New Delhi. He received his PhD from University of Wisconsin, Madison in 2006. He is a member of the R core group. His research interests include mathematical statistics, bioinformatics, statistical computing. He won the ASA 2004 Chambers award for Statistical Software.

Susan Thomas

Susan Thomas is Assistant Professor at the Indira Gandhi Institute of Development Research, Mumbai. She received her PhD in Econometrics from University of Southern California in 2017. Her research interests are in market microstructure, microfinance, financial econometrics. She is an organizer of the annual Emerging Markets Finance Conference.

About :

The conference seeks to bring together R users and data scientists from academia and industry to discuss and showcase applications of R in education, research and elsewhere. It is hoped that this will be a first in a series of large conferences involving R and data science, all over India and its vicinity.

A major goal of the conference is to make it feasible for R users in India and neighbouring countries to attend an International conference with top speakers and a world-class program. We want to provide a face-to-face welcome into the international R community to as many people as possible from all backgrounds and circumstances. We are committed to producing a conference that reflects the great diversity of the R Community.

Who should attend ?

We welcome R users with all levels of expertise from basic to advanced from academia and industry. Beginners will be able to gain an understanding of the power R and see a wide range of applications. Advanced users will have the opportunity to present talks by submitting abstracts and posters. All users will benefit from the tutorials.

Requirements :

The only requirement to benefit from this conference is that partcipants have familiarity with R and its package ecosystem. Attendees are expected to bring their own laptops to the conference for hands-on tutorials. Further instructions will be provided closer to the conference date.

Program :

This conference will be a a day-long, single-track event. The invited speakers are:

The morning session will include a tutorial and some talks. The afternoon session will consist of several talks of shorter and longer duration. This will be followed by poster session. A tentative schedule is listed below and is subject to change.

Registration :

The registration is done through the INDSTATS 2019 conference website which allows for registering selectively for the R@IISA conference. During registration, you will be asked to agree to the code of conduct.

Abstract Submission :

We invite you to submit abstracts for talks and/or posters using the Abstract Form. We also recommend that you register for R conference using the INDSTATS 2019 conference website using the same email address used for abstract submission. There is a registration fee, but it has been significantly reduced for participants from India and its vicinity. Note that if you abstract is accepted, you will be asked to register before you can participate.

Scholarships :

Thanks to our sponsors, we hope to provide between 40 and 50 scholarships for deserving participants from around India and its vicinity. We encourage, but do not require, applicants to present a poster or talk by submitting abstracts using the link above. Submitting an abstract will improve the chances of a scholarship award. In order to cross-reference abstract submissions with scholarship applications, we ask that the same email address be used everywhere.

The program committee will evaluate the applications and notify each applicant whether the application was successful or not.

Please note

Contact :

You may reach the organizers at r-iisa-2019@protonmail.com

For more details, visit R-conference page at https://r-iisa2019.rbind.io/