Dear CHANCE Colleagues,

Despite my beloved Red Sox winning the World Series, I wrote this column with a heavy heart. We were coming out of a week that saw a series of attempted mail bombings, a shooting that left two people dead at a Kroger grocery store in Kentucky, and another shooting at a Jewish synagogue in Pittsburgh that left 11 people dead. I learned that one of the victims of the Pittsburgh shooting was Joyce Fienberg, the widow of one of the founding co-editors of CHANCE, Stephen Fienberg.

Professor Fienberg initiated CHANCE and served as a co-editor with William F. Eddy from 1988–1991. Mrs. Fienberg was a mother, a grandmother, and a former research specialist at the Learning Research and Development Center at the University of Pittsburgh. Her friends described as a warm and caring person. These terrible events appear to be motivated by hate. Our thoughts are with the Fienberg family and the families of the other victims. Let us end the hate.

In this issue of CHANCE, we break away from our series of special issues and celebrate the fall with article diversity.

In our first article, and in the spirit of the election season, Arnold Barnett looks back at the 2016 election—one that shocked the prognosticators. He examines the analyses of two consolidators of the presidential polls, Real Clear Politics and FiveThirtyEight.com. He then evaluates conventional wisdom and arrives as an interesting conclusion regarding how much attention we should pay to such polls.

In 2003, Lucia de Berk, a Dutch pediatric nurse, was sentenced to life imprisonment for four murders and three attempted murders of patients in her care. She appealed in 2004, but was then convicted of seven murders and three attempts. Her conviction was controversial in the media and among scientists because quasi-statistical arguments played an important role in the case. The Supreme Court of the Netherlands reopened the case in 2008 as new facts were uncovered that undermined the previous verdicts. She was freed and her case was re-tried, resulting in exoneration in 2010. Richard Gill, Piet Groeneboom, and Peter de Jong evaluate the evidence in the case.

Jose Zubizarreta, Dylan Small, and Paul Rosenbaum discuss the concept of isolation using U.S. Fatality Analysis Reporting System (FARS) data as an example. They show us that a large, biased data system can contain a smaller, less-biased natural experiment.

In a summary of a JSM panel discussion about collaboration with non-statisticians, Daniel Mowrey, Jonathan Potts, Susan Spruill, Walter Stroup, and Michikko Wolcott highlight challenges primarily stemming from a lack of early involvement by statisticians in research projects and insufficient education about the interpretation of statistical evidence.

Briana Cameron, the 23andMe Research Team, and Robert Gentleman investigate and compare the responses to four surveys from the 23andMe website vs. a mobile app.

Full moons have been associated with myths and mystic phenomena. Could full moons be associated with motorcycle deaths? Donald Redelmeier and Eldar Shafir explore the lunacy of motorcycle mortality.

Deborah Dawson and Derek Blanchette provide a primer on meta-analysis, discussing challenging issues such as examining study heterogeneity, combining study results, bias, and quality assessment of the component studies and the meta-analysis.

In Taking a Chance in the Classroom, Chris Rohlfs discuss p-hacking and counts down 10 reasons not to drop insignificant regressors in model development.

Scott Evans

Back to Top