2018 Symposia

Research and Policy on Voter ID Laws and Voter Participation

Friday, February 16, 8:00-9:30am
Austin Convention Center, Ballroom G
Organizer: David Marker

This session brings attendees the most up-to-date science on voter suppression, the statistical impact of voter ID laws, and procedures for assuring that all eligible voters have a chance to vote. One speaker shares research into concerns about voter fraud and the impact of laws to prevent its occurrence. Another relays statistical expert testimony on voter ID laws in Texas, Pennsylvania, and Alabama. A third talk describes procedures states are implementing to maximize the chance that eligible voters will indeed be able to vote.

A Universe of Discoveries: Progress in Astronomy, Statistics, and Machine Learning

Friday, February 16, 10:00-11:30am
Austin Convention Center, Room 18A
Organizer: Chad Schafer

The fields of astronomy and cosmology have been revolutionized by a virtual flood of data gathered by instruments of unprecedented precision, and this growth is only continuing. For example, the Large Synoptic Survey Telescope, currently under construction in Chile, will gather 15 terabytes of data per day. It will cover over 1000 times the volume of current astronomical surveys, including approximately 40 billion astronomical objects, most of which are new discoveries. Comparable growth is ongoing or expected in other areas of astronomy. As data volumes grow, there is a corresponding growth in the complexity of the astronomical questions that could potentially be answered. Nevertheless, this potential will only be realized with the development and application of novel methods of data analysis specifically tailored to address the challenging issues raised by the high-dimensional, complex, and noisy data being gathered. These demands have given rise to a collection of researchers working at the interface of astronomy, statistics, and machine learning. Panelists are experts in this domain, and explore different aspects of the challenges their work addresses.

Advanced Data Analysis Techniques for Understanding Brain Function

Friday, February 16, 1:30-3:00pm
Austin Convention Center, Room 12A
Organizer: Rob Kass

This session features examples of the ways cutting-edge statistical methods are contributing important new insights into brain function. Speakers provide an overview of the opportunities and challenges, and explore in greater detail how statistical data analysis is helpful in understanding the mechanisms of anesthesia and loss of consciousness. The session also demonstrates use of a brain-computer interface for illuminating how neurons work in the brain and providing better treatment options for paralyzed patients.

Section U Business Meeting

Friday, February 16, 7:30-9:30pm
Hilton Austin, Room 615 A

Statistical and Computational Challenges in Genomics and Precision Medicine

Saturday, February 17, 8:00-9:30am
Austin Convention Center, Room 18C
Organizer: Michael Epstein
The objective of precision medicine is to develop patient-specific treatment, customized based on the patient’s genetic, environmental, and clinical risk profiles. The design of such therapeutics should improve patient outcomes and limit adverse effects of unnecessary treatment. Personalized medical care requires extensive knowledge of the genetic risk factors involved in human disease. To address this, researchers are applying cutting-edge sequencing technology to quickly assess genetic variation across a patient’s genome, and are using improved computational and bioinformatics data to assess pathogenicity of patient-specific variants. These advances will greatly accelerate the use of genetic information in clinical decision-making and further refine appropriate therapies. However, many statistical and computational issues in precision medicine remain to be resolved. Novel methods for designing clinical trials that incorporate patient-specific pharmacogenetic profiles are needed. Further, new analytic problems will emerge as precision medicine matures and incorporates the wealth of data emerging from electronic medical records. This session tackles these issues and many others related to precision medicine.

Survey Data Collection: Theoretical and Practical Perspectives

Saturday, February 17, 10:00-11:30am
Austin Convention Center, Room 19B
Organizer: Frauke Kreuter

Survey data are essential to research in many disciplines, such as political science, epidemiology, public health, demography, and economics, where these data have been relied upon to answer questions about the state of society, effective interventions, and predictions of election outcomes. However, dramatically falling response rates, failed predictions in high visibility elections, and the rise of alternative data sources have damaged the credibility of data generated from surveys in the last decade. Researchers seem to have moved away from probability samples in favor of less-representative but more convenient non-probability samples. Can we still draw valid inference from non-probability surveys? Under which circumstances and with what methods is this possible? Different disciplines have dealt with non-probability samples in very different ways. This panel brings together some of these different disciplines as well as incorporates both theoretical and practical perspectives on survey data collection, to make suggestions for the broader scientific community on use of surveys.

Using Wearable Device Data to Analyze and Improve Physical Activity and Health

Saturday, February 17, 10:00-11:30am
Austin Convention Center, Room 19B
Organizer: Raymond Carroll

Wearable or implantable devices such as accelerometers that measure physical activity and inactivity have become popular as ways for individuals to monitor their own behavior. The cost of these devices is falling and they are increasingly being deployed in health studies. While wearable devices hold great promise, questions of how to analyze the resulting data abound. This session addresses questions such as: what do wearable devices actually measure well and what do they not measure well; specifically how well do they measure sedentary behavior and sleep; which metrics derived from these devices are most predictive of health outcomes; how combining wearable devices and self-report instruments can predict health outcomes more efficiently; and lastly, how can defining profiles make public health policy more meaningful to large populations?

Science and the Fair Administration of Justice

Saturday, February 17, 1:30-3:00pm
Austin Convention Center, Ballroom G
Organizer: Alicia Carriquiry

The emergence of DNA analysis as an effective forensic tool in the 1990s was a revelation, in that it permitted quantification of the degree of association between a crime scene sample and a suspect. It also highlighted the fact that most other forensic practices lacked the rigorous and widely accepted scientific foundations of DNA profiling. The accuracy of tools once believed to be infallible, including fingerprints, is now in question. Under the Obama administration, there was a push to develop the scientific and statistical underpinnings of different types of forensic practice by encouraging the scientific community to collaborate with forensic practitioners and the legal profession on improving the science behind forensics. There was also an attempt to address problems in the reporting and interpretation of the results of forensic analysis. In this session, speakers highlight the role of the scientific community in ensuring justice and describe a successful state-level program to facilitate the introduction of science in judicial proceedings. They address the importance of distinguishing between personal opinion and science, and the type of statements every forensic report or testimony should be expected to include.

Pre-Election Polling Uncertainty: An Interdisciplinary Analysis

Sunday, February 18, 1:30-3:00pm
Austin Convention Center, Room 17B
Organizer: Andrew Gelman

Pre-election polls are not quite as good as proponents may advertise, but they’re better than many think. U.S. polls were off by about 2 percent in the 2016 U.S. presidential election, and U.K. polls were off by about the same amount in the “Brexit” vote. The source of the biggest forecasting problems in these votes came from insufficient attention to polling uncertainty. Where does polling uncertainty come from? This session discusses the increasingly popular approach of multilevel regression and post-stratification, which unifies the analysis of sampling and non-sampling error. Researchers from political science, sociology, economics, and engineering, as well as practitioners from the political polling and consulting arena, share insights into ideas from statistical modeling, big-data computing, and social science research used in survey design and adjustment.