Data Science newsletter – April 14, 2020

GROUP CURATION: N/A

As North Carolina health officials battle the outbreak of the Coronavirus across the state, the state has tightly controlled access to information related to the pandemic, providing some data but refusing to share others, even when requested.

The North Carolina Department of Health and Human Services updates a dashboard of information related to the virus each morning, including the number of laboratory-confirmed positive cases, the number of people tested for the virus and the number of people who have been hospitalized and who have died.

More recently, NCDHHS added a breakdown of additional data to its dashboard. For instance, you can look at a breakdown of the number of confirmed cases across the state by race or ethnicity. But that breakdown doesn’t tell you how many people from each group have been tested overall.

Not having the results from the backlog of testing means state officials can’t get a real-time picture of how deeply the disease has penetrated the population and when people might have hope of resuming their lives and getting back to work, an expert warns.

I have had fringe contact with more epidemiology than usual as of late, for obvious reasons, and I do understand this is only one corner of the discipline. I don’t mean this as a complaint dump, because most of economics suffers from similar problems, but here are a few limitations I see in the mainline epidemiological models put before us:

1. They do not sufficiently grasp that long-run elasticities of adjustment are more powerful than short-run elasticites. In the short run you socially distance, but in the long run you learn which methods of social distance protect you the most. Or you move from doing “half home delivery of food” to “full home delivery of food” once you get that extra credit card or learn the best sites. In this regard the epidemiological models end up being too pessimistic, and it seems that “the natural disaster economist complaints about the epidemiologists” (yes there is such a thing) are largely correct on this count. On this question economic models really do better, though not the models of everybody.

6/ the key idea is we don’t actually need geo data. We just need to know who was in contact with whom during a 2-week span. Whether it was at the gym or on the bus doesn’t matter for our purposes.

And we don’t need a big dystopian database. Much of the data can stay on phones.

A new study has begun recruiting at the National Institutes of Health in Bethesda, Maryland, to determine how many adults in the United States without a confirmed history of infection with SARS-CoV-2, the virus that causes coronavirus disease 2019 (COVID-19), have antibodies to the virus. The presence of antibodies in the blood indicates a prior infection. In this “serosurvey,” researchers will collect and analyze blood samples from as many as 10,000 volunteers to provide critical data for epidemiological models. The results will help illuminate the extent to which the novel coronavirus has spread undetected in the United States and provide insights into which communities and populations are most affected.

The study will be conducted by researchers at the National Institute of Allergy and Infectious Diseases (NIAID) and the National Institute of Biomedical Imaging and Bioengineering (NIBIB), with additional support from the National Center for Advancing Translational Sciences (NCATS) and the National Cancer Institute (NCI), all parts of NIH.

“This study will give us a clearer picture of the true magnitude of the COVID-19 pandemic in the United States by telling us how many people in different communities have been infected without knowing it, because they had a very mild, undocumented illness or did not access testing while they were sick,” said Anthony S. Fauci, M.D., NIAID director. “These crucial data will help us measure the impact of our public health efforts now and guide our COVID-19 response moving forward.”

This commentary is based on my 20 years of experience helping students suffering under toxic PIs. Here are some suggestions for graduate programs and institutions to help protect the trainees and ultimately reduce the number of toxic supervisors.

After college students joined the swarm of 200 million daily Zoom users this semester, experts bashed the company over privacy breaches and concerns about data sharing with third parties. That prompted Zoom executives to start a three-month re-evaluation of the videoconferencing platform’s encryption and licensing.

But online learning and meeting apps are just one aspect of privacy in higher education brought to the fore by the Covid-19 pandemic. Some academics fear the spreading crisis will be used to justify accelerated growth in intrusive observation of faculty members and students, further eroding individual rights in the name of education and public health.

The pandemic will hasten a “race to the bottom to a surveillance society, with very little indication that it’s going to make people safer,” said Chris Gilliard, a professor of English at Macomb Community College, in Michigan, who studies privacy and digital policy.

But some places, such as tiny Iceland, Europe’s most sparsely populated country – pop. 364,134, broadly equivalent to the number of people in Tulsa, Oklahoma – may be better placed to deliver some types of coronavirus information, and even answers than most, at least in the short term, according to public health experts, international government officials and others involved in responding to the outbreak.

“The size of a place matters. It tracks with the number of introductions of the virus. It is no coincidence the places now doing (the best work) share this feature,” said William Hanage, an epidemiologist at Harvard University’s T.N. Chan School of Public Health.

An immune reaction gone wild seems to be linked with the most severe cases of pandemic Covid-19. Here’s what happens.

The Toronto-based Vector Institute for Artificial Intelligence is repurposing artificial intelligence computing infrastructure in order to support data-intensive analysis and decision-making for Ontario’s Pandemic Threat Response (PANTHR) project.

The Vector Institute had already purchased 240 RTX6000 graphics processing units, which are designed for multiple high-performance computing applications, including deep learning. The hub says it will repurpose this technology to allow researchers to process large amounts of data and allow for deeper insights into the current pandemic. The institute is also planning to provide data science expertise as part of this initiative.

PANTHR was created by the provincial government due to the current need to analyze and better detect, plan, and respond to the COVID-19 pandemic. As part of the project, the province has committed that PANTHR will free up access to de-identified, integrated data on publicly-funded administrative health services records. The health platform will additionally contain clinical data from special registry collections, such as the Critical Care Information System, which reports on critical care capacity in the province.

Artificial intelligence (AI) is evolving—literally. Researchers have created software that borrows concepts from Darwinian evolution, including “survival of the fittest,” to build AI programs that improve generation after generation without human input. The program replicated decades of AI research in a matter of days, and its designers think that one day, it could discover new approaches to AI.

“While most people were taking baby steps, they took a giant leap into the unknown,” says Risto Miikkulainen, a computer scientist at the University of Texas, Austin, who was not involved with the work. “This is one of those papers that could launch a lot of future research.”

If you’re like me, you’ve been watching the daily data on the coronavirus pandemic, seeking glimmers of hope in the trajectories: the infected, the hospitalized, the intubated, the dead.

If only there were more understanding to be had. The more I look at the numbers, the more I see their flaws. Here are my top 10.

1. The number of infected is close to meaningless. Only people who get tested can be counted, and there still aren’t enough tests — not even close, and not in any country save perhaps Iceland. The best we can do is estimate how many people are sick by guessing what percentage of the infected can obtain a test. In the U.S., for example, anecdotal evidence suggests that people need to be ill enough to be hospitalized. About 10% of cases merit hospitalization, so the actual number of infected might be about ten times larger than what’s reported.

The new BMJ data is hugely important as the majority of new information and findings released daily worldwide is from the potentially small proportion of people who have shown symptoms, sought hospital help, undertook a test and tested positive. This is different to previous epidemics such as SARS, where most of the infections were symptomatic and could be traced.

Ultimately, widespread antibody testing, which is still not imminent, will be able to tell us how many people have already had COVID-19. This will give a better approximation of the total number of infections. This will be important in making decisions on lifting social distancing measures.

For example, if antibody testing suggests that a large proportion of the population has had COVID-19 already, there is a smaller chance of asymptomatic and undiagnosed cases spreading the infection once restrictions are lifted. But if only a very small proportion of the population has had the infection, then lifting of social distancing measures may have to be delayed until vaccination strategies are ready to be implemented.

Boston College will begin recruiting the first class for its highly anticipated Human-Centered Engineering program, a major that will integrate BC’s core liberal arts focus and a rigorous engineering curriculum to prepare students to find solutions that address critical human needs.

Launching in the fall of 2021 and guided by the fundamentals of engineering education, the Human-Centered Engineering program will emphasize experiential learning and application-based course offerings, giving students opportunities to work in collaborative teams and across disciplines to address critical issues in the areas of the environment, health, and energy.

Aligned closely with the University’s new Schiller Institute for Integrated Science and Society, the program will create new avenues for faculty and student research in the applied sciences and respond to the high level of interest in engineering among prospective applicants considering majors in STEM fields.

Apple Inc. responded to Democratic Senators who sent a letter to Chief Executive Officer Tim Cook with questions related to the privacy of the iPhone maker’s Covid-19 screening tools.

In a letter dated Friday, Apple said the tools “were built to protect the privacy and security of users’ data.” The company also answered questions related to data sharing, agreements with government agencies, and the accessibility of the tools.

Online May 30. “We are looking for 15 minute talks (give or take a minute) on topics relevant to the R community, followed by moderated Q&A for about 5 minutes. As we go virtual this year, we are seeking snappy and eclectic content; it’s okay if your talk errs on the shorter side. It is also fine if you want to touch on a few different, but related, topics (e.g. three related functions from the same package), or if you wish to reuse prior material (e.g. blog post, package vignette, etc.).” Deadline for submissions is April 30.

“TLDR: Most visualization design and evaluation methods don’t explicitly consider beliefs. Applying a Bayesian framework to visualization interaction provides a more powerful way to diagnose biases in people’s interactions with data, like discounting or overweighting data in judgments or decisions. We can also use Bayesian models of cognition to evaluate visualizations that present uncertainty, to personalize how we visualize or explain datasets, and to predict different individuals’ future responses to data.”

“In March 2020, an ACM Presidential Task Force was formed to provide quick advice to conference organizers suddenly facing the need to move their conference online in light of the social distancing recommendations and global restrictions on travel due to the COVID-19 pandemic. We provide concrete advice for events of all sizes. We discuss the tasks required of organizers, specific platforms that can be used and financial considerations. We collect examples of conferences that have gone virtual and lessons learned from their experiences.”

Google Cloud’s AI Platform recently added support for deploying TensorFlow 2 models. This lets you scalably serve predictions to end users without having to manage your own infrastructure. In this post, I’ll walk you through the process of deploying two different types of TF2 models to AI Platform and use them to generate predictions with the AI Platform Prediction API. I’ll include one example for an image classifier and another for structured data. We’ll start by adding code to existing TensorFlow tutorials, and finish with models deployed on AI Platform.

“With a third of the world population under lockdown to prevent the spread of the virus, current containment measures look set to be in place for the majority of 2020. The longer the pandemic goes on, the more important social science and social scientists become in managing the social, political, economic and cultural upheaval that COVID-19 has thrust upon us all. Our journals team have built a free to access microsite featuring the latest medical research into COVID-19 published on SAGE Journals but also social and behavioral science insights into working, living and educating during a pandemic, effects on national and international infrastructure and how best to manage stress and anxiety.”