“Cox and Efron’s techniques are used on a daily basis in the practice of statistical science, and have made an enormous impact in all the sciences which rely on the analysis of data,” the jury’s citation said.

Cox’s contribution, “the Cox regression,” is a powerful tool to explain the duration of a time interval between two events of interest, which depends on identifiable factors rather than mere chance (for instance, the mortality of a group of individuals due to a particular disease or a risk factor like environmental pollution). It finds use in such varied fields as cancer research, epidemiology, economics, psychology or sociology, and even in the testing of the resistance and durability of industrial products. The jury illustrates the technique’s application in the medical field by citing the conclusion that even a year of smoking cessation contributes to reduce mortality.

Bradley Efron, Stanford University, meantime, is the inventor of the bootstrap, a “deceptively simple” technique, as the jury terms it, to estimate the margin of error of a given outcome; a must-know in science, without which results are worthless.

Both contributions date from decades ago and both laureates found it hard to pick just one out of the multiple applications found since then. David Cox, University of Oxford, declared himself “enormously surprised and gratified” by the sheer range of scientific problems his method has helped address. Cox’s technique, published in 1972, is now the second most cited statistics paper in modern scientific literature.

Cox’s move into statistics was motivated by the military imperatives of the aeronautics industry in the Second World War. Efron, who met Cox in London in 1972, had been nudged towards statistics by his father’s love of mathematics and sport. He says part of what led him to the bootstrap technique, published in 1979, was a conversation he had with Cox then about another statistical analysis method.

The two laureates concur that their own methods, and statistical tools in general, will become increasingly necessary in the practice of science, more reliant by day on the analysis of massive data sets.

*Abridged from* http://www.fbbva.es/TLFU/tlfu/ing/microsites/premios/fronteras/galardonados/2016/ciencias.jsp

We are delighted to welcome to the *Bulletin’*s team of Contributing Editors the following “new faces”:

**Yoram Gat** (Google Israel)

**Takis Konstantopoulos **(Uppsala University, Sweden)

**Regina Nuzzo **(Gallaudet University, Washington DC, USA)

**Kavita Ramanan** (Brown University, RI, USA)

They will join the existing members of the team, Anirban DasGupta, David Hand, Xiao-Li Meng, Dimitris Politis and Terry Speed.

Look out for Anirban’s **Student Puzzle Corner**—back after a hiatus—and Xiao-Li’s **XL-Files (with a prize!)**. Next issue we’ll bring you columns from Yoram Gat and Terry Speed, and maybe more…

Don’t forget, if you have a **POPI** (a **Project, Object or Perspective of ****Interest**), send us a note at bulletin@imstat.org.

Fox is among 102 scientists and engineers (only 19 via the National Science Foundation) who are being recognized by the White House for advancing the frontiers of science and technology and serving the community through scientific leadership, public education, and community outreach.

“I congratulate these outstanding scientists and engineers on their impactful work,” said President Barack Obama in a press release in January announcing the winners. “These innovators are working to help keep the United States on the cutting edge, showing that Federal investments in science lead to advancements that expand our knowledge of the world around us and contribute to our economy.”

PECASE winners are chosen from among nominees submitted by a dozen federal agencies and the intelligence community for making significant contributions to America’s continuing leadership in science and technology. Fox was nominated for the award by the National Science Foundation for her “groundbreaking work in large-scale Bayesian modeling and computational approaches to time series and longitudinal data analysis, and for outstanding outreach and mentoring of women in computer science and statistics.”

In addition to her Amazon Professorship (with appointments in Statistics and CSE), Fox is a Data Science Fellow in UW’s eScience Institute and co-created the UW’s Coursera specialization in machine learning in collaboration with UW CSE professor Carlos Guestrin. The PECASE is the latest in a long list of honors for Fox, who is the recipient of a Sloan Research Fellowship, ONR Young Investigator Award, NSF CAREER Award and the MIT EECS Jin-Au Kong Outstanding Doctoral Thesis Prize, among many others. Fox’s research has been applied in a wide range of domains, including neuroscience, finance and econometrics, social networking, and more.

]]>Voting opens soon—look out for the email with your personalized link—and closes June 16.

]]>]]>

The medal selection committee acknowledged that Professor Scott is a world leader in survey sampling theory and analysis of case control studies. His methods are used in many applications and he has also contributed substantially to research in public health. His work has particular relevance to obtaining reliable data from sampling, developing effective and simple methods that can take account of survey design features and deal with missing data.

His 1981 paper on categorical survey data was recognized as one of the 19 landmark papers in survey sampling by the International Association of Survey Statisticians in their 2001 Centenary volume. These methods, developed with Professor Rao, called Rao–Scott adjustments, are widely used and incorporated in several software packages for survey data analysis. In addition to developing a large body of novel and important statistical methodologies, he has been an advisor to official agencies nationally and internationally.

On receiving the Medal, Professor Scott said: “I feel very honoured to receive the Jones Medal named in honour of [New Zealand’s] most celebrated mathematician, Sir Vaughan Jones,” recalling that he had taught Jones in 1972.

Scott is a Fellow of the RSNZ, ASA, IMS and the Royal Statistical Society. He is an Honorary Life Member of the New Zealand Statistical Association and received its premier award, the Campbell Prize in 2012. In 2006 he received the ASA/SSC Waksberg Award for outstanding contributions to survey methodology.

]]>—

**INTRODUCTION**

A century ago, Einstein and others laid down the foundation of what is now known as general relativity — the theory of how mass and the curvature of spacetime interact. One prediction of this theory is the radiation of energy through gravitational waves (Einstein 1916) from accelerating massive systems like astronomical binaries. While normally an immeasurable effect for widely separated, non-compact systems like the Solar System, close binaries with black hole or neutron star components emit potentially detectable and distinct gravitational-wave signatures. Detection by laser interferometers, like the kilometer-scale instruments in Livingston, Louisiana and Hanford, Washington, has been pursued because they are capable of detecting the diminutive effect of passing waves—in the detections discussed here, the length of the 4 km interferometer arms varied by only ~10^{−18}m. Most searches for gravitational waves from compact binaries use models which are based on general relativity and parameterized by the physical properties of the binary (e.g., compact object masses, angular momenta, orientation and position; usually more than 15 parameters). These models describe the effect of the wave impinging on a network of gravitational-wave detectors. To account for modeling uncertainty, or when a complete signal model is unavailable, more generic searches are employed, which assume no particular signal morphology. In the latter half of 2015, the two LIGO interferometers (Abbott et al. 2016e) recorded several transient events from the mergers of binary black holes (Abbott et al. 2016b), confirming their existence and measuring their properties (Abbott et al. 2016a) using a variety of sophisticated statistical techniques. Abbott et al. (2016g) describes the basic, order-of-magnitude physics behind the radiated signal and its interpretation in the first detection.

The second detection of GW151226 and weaker, but still compelling, LVT151012 have left no doubts in the minds of scientists that gravitational waves exist and match with our description of Nature to high precision. The detections also provided a unique opportunity to test general relativity in its position as the prevailing theory of gravity. The vast majority of alternative theories of gravity predict gravitational waves, but with modifications according to their peculiarities. The searches and most of the parameter estimation use models which assume general relativity (GR) is the correct description of the interaction between gravity and matter. The generic transient analyses mentioned previously do not require GR to be entirely accurate, but still require most of its fundamental tenets. There are also a variety of explorations into systematic deviations from GR—a summary of LIGO–Virgo Collaboration work is in Figures 7&8 in Abbott et al. (2016b). One of many efforts to translate the observations into constraints on various GR extension theories can be found in Yunes et al. (2016).

The community response to the discovery has been nearly ubiquitously positive and constructive—they are incorporating and building on our result. There have been very few serious, peer-reviewed attempts at debunking the result. The statistical arguments remain unchallenged in the literature, but some have questioned various pieces of the experiment and its interpretation. For example, in Chang et al. (2016), it is argued that a part of our frequency response to gravitational waves was flawed. However, no counterclaim has yet been considered to be credible.

**CONTEXT: ASTROSTATISTICS AND GRAVITATIONAL-WAVE DETECTION**

The field of astrostatistics is quite rich, and the LIGO–Virgo Collaboration has taken advantage of a solid foundation forged by others before us. The literature has many earlier examples of statistical modeling and its use with large data sets, weak signal strength, possible measurement biases, and disentangling mixtures of populations: for example, exoplanet detection with Kepler data (e.g. see work by Foreman-Mackey, Hogg, Rogers, and many more). Another very concrete, though not yet realized, example is the LSST (Large Synoptic Survey Telescope) project ** ^{1}**, slated to begin in the 2020s. They will be dealing with an event rate in the tens of thousands

The transient searches employed by LIGO can be separated into those which assume a particular source model (e.g. searches for binary mergers, see Abbott et al. (2016b) and references therein) and more generic searches which only enforce reconstructed signal consistency between instruments (see Abbott et al. (2016c) and references therein). Many of the statistics involving the matched filtering were derived in the early 2000s: both types of searches use time-series filtering algorithms (Anderson et al. 2001; Allen et al. 2012). These algorithms produce lists of times and amplitudes relative to the noise (encoded in the signal-to-noise ratio, or SNR) which characterize putative signals embedded in a noisy data stream. The stochastic nature of the noise means that the SNR has a statistical distribution with larger SNRs becoming increasingly improbable.

The main function of statistics like the SNR and χ^{2} (Allen 2005) is to distinguish the event candidate from the non-astrophysical transient background distribution and establish its statistical significance. Searches must also deal with transients in the data which are induced by the instrument and its environs (Abbott et al. 2016a) — colloquially called “glitches”. This additional population imposes fatter tails on the idealized SNR distribution. In practice, there is no analytic description for the non-astrophysical environmentally-induced transient SNR distribution, and so this must be measured empirically.

**PARAMETER ESTIMATION**

Once an event time of interest is observed by the searches, a posterior distribution over the parameters describing the source of a signal is sampled using forward-modeling Bayesian methods that demand explicit modeling of signal and noise alike. The likelihood function is formed from the residuals left after signal subtraction from the data time series. While this is a (mostly) straightforward application of Bayes’ Law, there is no analytical marginalization available, so stochastic sampling techniques, particularly Markov Chain Monte Carlo and nested sampling (Abbott et al. 2016a; Veitch et al. 2015) are employed. This likelihood, coupled with the priors on source parameters (e.g., uniform in compact object masses and spins, isotropic in orientation, uniform in volume in the local universe), provide the posterior probability density for the source parameters.

Another analysis has been developed to account for possible glitch behavior at the time of a signal and to mitigate uncertainties in the signal model. The non-Gaussian components of the noise (assumed uncorrelated between instruments) can be modeled simultaneously with the signal (assumed correlated across detectors). The reconstruction is obtained using a reverse-jump Markov Chain Monte Carlo (RJMCMC) ** ^{2}**. This enables jumps between model spaces in addition to traditional MCMC jumps within a single model space, allowing for model comparison to be done “on the fly” (Cornish & Littenberg 2015; Abbott et al. 2016c). It is a powerful tool in unparameterized signal reconstruction and glitch rejection, but this method has relaxed assumptions about the astrophysical signal relative to the rigorous parameterized modeling imposed by the previously mentioned MCMC and nested sampling techniques, and is not typically used to generate posteriors over the physical properties of the binary.

**RATES AND MASS DISTRIBUTIONS**

One of the key science outputs from these observations is the density of merging binary black holes in spacetime (the merger rate), expressed as a number of mergers per cubic gigaparsec (Gpc^{3}) per year ** ^{3}**. To calculate this number, the Collaboration must calculate the number of detected mergers (the “numerator” in units of counts) and the spacetime volume that it has searched (the “denominator” in units of Gpc

While two *confident *gravitational-wave detections were reported (GW150914 and GW151226) there was a third candidate (LVT151012) identified with measured cumulative accidental background event coincidence (or false alarm) rate of (2*.*3 yr)^{−}^{1} (Abbott et al. 2016d). This candidate has a significance, or *p*-value, of 0*.*045 (Abbott et al. 2016b). To account for our uncertainty about the origin of this trigger we fit a mixture model comparing the SNR distributions of both the terrestrial and astrophysical populations to the set of triggers from our searches, including the three candidates and many more (Abbott et al. 2016b,i,h; Farr et al. 2015). From this model, we infer a posterior probability that the LVT151012 trigger is associated with an astrophysical source of 0*.*86. We use the posterior on the astrophysical population in the mixture model to infer coalescence rates.

Fitting a hierarchical power-law model to our three candidates, accounting for our mass measurement uncertainties and selection effects (Loredo 2004; Mandel et al. 2016) yields a posterior median and 90% credible interval of *α*=2.5^{+1.5}_{−1.6} (Abbott et al. 2016b); not surprisingly, given three possible detections, the source population is poorly constrained. A full accounting of rates under different population assumptions can be found in Abbott et al. (2016b), but the union of the rates estimate provides a conservative 90% credible interval of 9–240 Gpc^{−3 }yr^{−1}.

**OUTLOOK**

The posterior predictive distribution for the rate can be used to extrapolate the expected number of future detections over the next set of observational runs (labeled O2, O3, etc.). While dependent on progress with planned improvements in the instruments (Abbott et al. 2016e,c), using a reasonable range of estimates for the expected detector sensitivities in the upcoming six-month LIGO observing run from late 2016 (Abbott et al., 2016f), the probability of at least 10 more confident detections (like GW150914 and GW151226) in the next run is between 15–80% (see Fig. 13, Abbott et al., 2016b).

The measurement of the binary parameters, their statistical significance, and occurrence rate are definitive statements: our methods are well tested from previous data-taking runs, and the evidence presented is a clear affirmation of the significance of the detections. However, as O2 evolves and more events are collected, the rates will become better measured. As a consequence, several groups inside and outside the Collaboration are working on more sophisticated hierarchical models in attempts to incorporate more astrophysics and do better, more nuanced model selection. We seek the answers to questions like: “How do heavy black holes form?” and “What are their environments and interactions?”. Speaking more directly to physics, we’ve confirmed yet another (elusive!) piece of general relativity, and pushed the frontiers of precision science and large-scale experimental physics. The astronomy community has also been very excited: another potential discovery in the next few years may be a neutron-star black-hole binary. Should this happen, there is a chance we can correlate the gravitational-wave signal with a flash of gamma rays (very high energy light). This would allow for a rich laboratory of science involving the high-density matter composing neutron stars, otherwise inaccessible in terrestrial laboratories. Beyond binary merger events, there is potential for detections of supernova, semi-monochromatic GW emission from deformed neutrons stars, and the gravitational-wave analog to the cosmic microwave background. The Collaboration’s plan for just the next year fill many pages (Collaboration 2016).

GW150914, in and of itself, did not directly affect humanity. Indeed, all gravitational-wave events pass by without notice by the population at large. However, the technological and sociological benefits of the Collaboration’s journey to discovery of gravitational waves are immeasurable. As a beginning, we have developed an instrument capable of extremely high precision measurements and we’ve developed leading edge data analysis algorithms and statistical modeling techniques. Furthermore, the Collaboration has provided data releases to demonstrate to society the fruits of endeavor that their investment has directly contributed to. About an hour of the time series data surrounding each event has been released by the LIGO Open Science Center (LOSC), free for exploration and accompanied by basic tutorials in their use** ^{4}**. A full data release for O1 will occur, but may take several more months to assemble. In the meantime, those interested can obtain the data from the fifth and sixth science runs (2005–2010) from the LOSC site above.

Maybe less concretely, but more poignantly: we’ve allowed the world to “listen” to black holes for the first, and assuredly not the last, time.

—

FOOTNOTES

^{1 }Large Synoptic Survey Telescope: https://www.lsst.org/ ↩

** ^{2}** For a good introduction, see https://www-sigproc.eng.cam.ac.uk/foswiki/pub/Main/SJG/hssschapter.pdf ↩

** ^{3}** One parsec is 3

** ^{4}** https://losc.ligo.org/events/ ↩

**REFERENCES**

Abbott, B. P., Abbott, R., Abbott, T. D., Abernathy, M. R., Acernese, F., Ackley, K., & Adams, C. 2016a, *Classical and Quantum Gravity*, 33, 134001

Abbott, B. P. et al. 2016b, *Phys. Rev. D*, 93, 122003

——. 2016c, *Phys. Rev. D*, 93, 122004

Abbott, B. P. et al. 2016a, *Physical Review Letters*, 116, 241102, 1602.03840

——. 2016b, *Physical Review X*, 6, 041015, arXiv:1606.04856

——. 2016c, ArXiv e-prints, arXiv:1602.03845

——. 2016d, *Phys. Rev. D*, 93, 122003, arXiv:1602.03839

——. 2016e, *Physical Review Letters*, 116, 131103, arXiv:1602.03838

——. 2016f, *Living Reviews in Relativity*, 19, arXiv:1304.0670

——. 2016g, ArXiv e-prints, 1608.01940

——. 2016h, ArXiv e-prints, arXiv:1606.03939

——. 2016i, ArXiv e-prints, arXiv:1602.03842 Allen, B. 2005, Phys. Rev. D, 71, 062001

Allen, B., Anderson, W. G., Brady, P. R., Brown, D. A., & Creighton, J. D. E. 2012, *Phys. Rev. D*, 85, 122006

Anderson, W. G., Brady, P. R., Creighton, J. D. E., & Flanagan, E. E. 2001, *Phys. Rev. D*, 63, 042003

Chang, Z., Huang, C.-G., & Zhao, Z.-C. 2016, ArXiv e-prints, 1612.01615

Collaboration, 2016, *The LSC-Virgo White Paper on Gravitational Wave Searches and Astrophysics (2016-2017 edition)*, Tech. rep. https://dcc.ligo.org/T1400054/public

Cornish, N. J., & Littenberg, T. B. 2015, *Classical and Quantum Gravity*, 32, 135012, 1410.3835

Einstein, A. 1916, *Sitzungsber. Preuss. Akad. Wiss. Berlin (Math. Phys.)*, 1916, 688

Farr, W. M., Gair, J. R., Mandel, I., & Cutler, C. 2015, *Phys. Rev. D*, 91, 023005, arXiv:1302.5341

Foreman-Mackey, D., Hogg, D. W., & Morton, T. D. 2014, *The Astrophysical Journal*, 795, 64

Hogg, D. W. 1999, ArXiv Astrophysics e-prints, arXiv:astro-ph/9905116

Loredo, T. J. 2004, in *American Institute of Physics Conference Series*, Vol. 735, ed. R. Fischer, R. Preuss, & U. V. Toussaint, 195–206, arXiv:astro-ph/0409387

Mandel, I., Farr, W. M., & Gair, J. 2016, *Extracting distribution parameters from multiple uncertain observations with selection biases*, Tech. Rep. P1600187, LIGO, https://dcc.ligo.org/LIGO-P1600187/public

Veitch, J. et al. 2015, *Phys. Rev. D*, 91, 042003, 1409.7215

Yunes, N., Yagi, K., & Pretorius, F. 2016, *Phys. Rev. D*, 94, 084002

—

After a long gap, we now resume the problem corner, and it is the turn of a problem on probability this time. The problem is at an interesting intersection of probability, analysis, and number theory.

Imagine that you are tossing an honest die repeatedly, and your score after the $n$th roll, say $S_n$, is the sum of the first $n$ rolls. This, of course, is an integer between $n$ and $6n$. Will $S_n$ ever be a prime number for some $n$? For infinitely many $n$? What can we say about how many rolls does it take for $S_n$ to be a prime number for the first time? Does it take just a few rolls? Is the expected waiting time finite? Can we give an approximate value for the expected waiting time? And so on.

Here is the exact problem of this issue:

Let $X_1, X_2, \cdots $ be iid discrete uniform on the set $\{1, 2, \cdots , 6\}$, and let for $n \geq 1, S_n = \sum_{i = 1}^n X_i$. Let $\mathcal{P}$ denote the set of prime numbers $\{2, 3, 5, 7, \cdots \}$, and $\tau = \inf \{n \geq 1: S_n \in \mathcal{P}\}$.

(a) Is $P(\tau < \infty ) >0$?

(b) If $P(\tau < \infty ) >0$, does it have to be $1$?

(c) Show that $E(\tau ) > \frac{7}{3}$.

(d) Is $E(\tau ) < \infty $?

(e) If $E(\tau ) < \infty $, give an approximate numerical value for it.

(f) Conjecture if the variance of $\tau $ is finite.

(g) Is $P(S_n \in \mathcal{P} \, \mbox{for

infinitely many}\, n) = 1$?

*Note: Answer as many parts as you can; do not be disappointed if you cannot answer all the parts.*

Contributing Editor Anirban DasGupta writes:

The problem asked was to derive an asymptotically correct $100(1-\alpha )\%$ confidence interval for $F(\mu )$, given an iid sample $X_1, \cdots , X_n$ from a distribution with CDF $F$, finite mean $\mu $ and variance $\sigma ^2$, and a density $f(\mu )$ at $\mu $, in the sense that $F$ is differentiable at $\mu $ with a derivative $f(\mu )$. The mean and the variance are considered unknown, and no functional form of $F$ or $f$ is assumed.

The problem is not entirely simple; there is some literature on it. If we define the empirical process $G_n(t) = \sqrt{n}\,[F_n(t)-F(t)],$ where

$F_n(t) = \frac{1}{n}\,\sum_{i = 1}^n\,I_{X_i \leq t}$, then consider the decomposition $\sqrt{n}\,[F_n(\bar{X})-F(\mu )] = [G_n(\bar{X}) – G_n(\mu )] + \sqrt{n}\,[F_n(\mu ) – F(\mu )] + \sqrt{n}\,[F(\bar{X}) – F(\mu )]$.

By using the multivariate central limit theorem, the delta theorem, and the order of the oscillation of the empirical process $G_n(t)$ in small intervals, one can show that $\sqrt{n}\,[F_n(\bar{X})-F(\mu )] \stackrel {\mathcal{L}} {\Rightarrow } N(0, V(F))$, where $V(F) = F(\mu )(1-F(\mu )) + \sigma ^2\,f^2(\mu ) + 2\,f(\mu )\,E_F[(X-\mu )I_{X \leq \mu }]$. Construction of a confidence interval for $F(\mu )$ requires consistent estimation of $F(\mu )$, $\sigma ^2$, $E_F[(X-\mu )I_{X \leq \mu }]$, and $f(\mu )$. The first three are easily done. Consistent estimation of $f(\mu )$ can be done by using various standard density estimation methods, but because $\mu $ is considered unknown, the assumptions on $f$ are stronger than what one needs for pointwise consistent estimation of $f(x)$ at any known $x$. Alternatively, one can bootstrap $\sqrt{n}\,[F_n(\bar{X})-F(\mu )]$, and find bootstrap estimates of the variance or directly find quantiles of the bootstrap distribution.

]]>*“How could that happen?”* was perhaps the question of the year for 2016. Other than a small percentage of perceptive minds, which I hope include disproportionately more of my fellow statisticians, the rest of the human population seems to still be coping with the aftermath (*afterstat*?) of 2016. Forever, 2016 will remain in our memory as an *extraordinary* year, literally. It was a year that also saw the departures of more extraordinary figures than in any other year in my memory, although I am acutely aware of my age-induced ability to create alternative facts. The departure of the daughter–mother pair, Carrie Fisher and Debbie Reynolds, on two consecutive days just before the departure of 2016 itself, sadly dramatized “The Year of The Reaper,” as *Time* put it (on its cover).

And although we are still years away from statisticians sharing a *Time* cover with Hollywood celebrities, our profession did have its heavy share of the Year of The Reaper. We started the year with the shocking departure of one of the most prolific, and kindest, scholars of our time, **Peter Hall** (11/20/1951–1/9/2016). Only a month after, we lost one of the most engaging and forceful pioneers, **Emanuel Parzen** (4/21/1929–2/6/2016). Less than 12 weeks later, we had to bid farewell to the legendary advocate and nurturer of statistics and statisticians, **Ingram Olkin** (7/23/1924–4/28/2016). Six weeks on, we lost another eminent and penetrating scholar, **Vidyadhar Godambe** (6/1/1926–6/9/2016). In early fall, we learned of the demise of **Theodore Anderson** (6/5/1918–9/17/2016), a renowned pioneer bridging statistics and econometrics. Ten weeks later the statistics community was shaken again by the departure of his contemporary, **Charles Stein** (3/22/1920–11/24/2016), a towering statistical intellect for all times. The end of the year came with the sad news of the passing of **Stephen Fienberg** (11/27/1942–12/14/2016), another prolific scholar, as well as an energizing leader, of our profession.

Sadly, this list is far from exhaustive, especially if we include scholars in closely related fields. Just locally, in April, I attended a very moving memorial event for **James Ware** (10/27/1941–4/26/2016), a leading figure, both scholarly and professionally, in biostatistics. And in July, Harvard mourned the loss of **Howard Raiffa** (1/24/1924–7/8/2016), a giant in game theory and decision analysis (and a founding uncle of Harvard’s Statistics Department).

Of course, we all will enter history sooner or later. The differences are that each of us may belong to a different cell of a 2×2×2 table: those who make history or not; those who care about doing so or not; and those who will be remembered by history or not. Regardless of which cell represents an ideal life and which cell is *my* destiny, the inspirations generated by such leading scholars and scholarly leaders can enrich our “cell life” in multiple ways. Therefore, the year of 2016 should be remembered also as a particularly rich year of inspiration. And to help the younger (than me) generations to remember all of them, and other giants of our beloved profession, I will leave you to correctly identify all nine of them in the pictures below, as well as those others in the pictures, including one whose centennial recently passed. The first eight people who can correctly identify everyone in the pictures (and who are above 21) will be invited to a “libation and inspiration” gathering at JSM 2017 to celebrate the lives of these nine (and hopefully not many more). [Email your entry and age-proof to meng@stat.harvard.edu*.*]

**How many of these 14 people can you name? There’s a prize for the first 8 to tell Xiao-Li who they all are!**

]]>

The Applied Public Health Statistics Section of the American Public Health Association (APHA) invites nominations for the 2017 Mortimer Spiegelman Award, which honors a statistician below the age of 40 in the calendar year of the award who has made outstanding contributions to health statistics, especially public health statistics. The award was established in 1970 and is presented annually at the APHA meeting. The award serves the following three purposes:

1. To honor the outstanding achievements of both the recipient and Spiegelman

2. To encourage further involvement in public health by the finest young statisticians

3. To increase awareness of APHA and the Applied Public Health Statistics Section in the academic statistical community

The Spiegelman Award recipient must be a health statistician who has made outstanding contributions to statistical methodology and its applications in public health (broadly defined).

The award is open to early career investigators regardless of race, gender, sexual orientation, nationality or citizenship. Specifically, candidates must meet at least one of the following criteria:

• Candidate must be under age 40 throughout the award calendar year; or

• Candidate must have obtained a terminal degree in statistics or a statistics-related field in the last 10 years.

For those receiving a terminal degree after considerable professional experience, contributions during and subsequent to the degree will be considered by the committee, and nominators are strongly encouraged to contact the Committee Chair with any questions about whether the nominee qualifies as an early career investigator.

Please email a nominating letter that states the candidate’s date of birth and how their contributions relate to public health concerns, up to three letters of support, and the candidate’s CV to the award committee chair, Tyler VanderWeele, at tvanderw@hsph.harvard.edu. Nominations are due by April 1, 2017.

Nominations are being sought for the 2017 National Institute of Statistical Sciences’ (NISS) Jerome Sacks Award for Outstanding Cross-Disciplinary Research. The prize recognizes sustained, high-quality, cross-disciplinary research involving the statistical sciences.

An award of $1,000 will be presented during the NISS reception at the Joint Statistical Meetings (JSM) in Baltimore, July 29–August 3, 2017.

For more information, including a list of previous award winners, please see: www.niss.org/about/awards/jerome-sacks-award-outstanding-cross-disciplinary-research

To nominate an individual, submit as one PDF document the following information to sacksaward2017@niss.org by May 1, 2017: a nomination letter (maximum two pages); supporting letters from two individuals (other than nominator); and the nominee’s CV.

Questions about the award or the nomination process can be sent to the email address given above.

]]>IMS President 2016–17 Jon A. Wellner, University of Washington, writes:

**History: Continuity and Discontinuity**

In preparing to write this article for the *Bulletin*, I took the opportunity to read several of the introductory articles written by previous IMS Presidents over the past 10 years. These pieces convey a wonderful sense of enthusiasm for the mission and goals of the IMS. They also contain considerable wisdom concerning the challenges and cross-winds the IMS has faced and continues to navigate in continuing to serve as a meeting place and society for those interested in scientific research and the effective application of probability and statistics in the modern world.

In reading these articles, I was struck by both the continuities and the discontinuities in the themes raised. Many of the new Presidents have championed one or more new initiatives, but each President has only one year to both introduce such initiatives and to follow through with implementation. In short, it is difficult to achieve continuity at the level of the President. On the other hand, other members of the IMS Executive Committee serve three-year terms and it is not uncommon for these members to serve two consecutive terms. Similarly, our current practices concerning editors and many of our standing committees involve three-year terms (or sometimes longer). An important source of continuity has been in our paid staff, the Executive Director Elyse Gustafson, and *IMS Bulletin* Assistant Editor Tati Howell. Elyse began as IMS Executive Director in 1997 and serves as the major source of institutional memory for our society. Tati Howell has been contracted to work with IMS since 2002. We owe them both a huge debt for keeping the IMS on an even keel over a long period of time. I personally owe them both for important assistance on many occasions!

In the remainder of this article my goal is to give a brief snapshot of some of the current activities of the IMS, including our journals, meetings, changes (e.g., the winding down of *CIS*) and opportunities for new activity, as well as reminders about some recent procedural changes.

**IMS Executive Committee**

As noted above, one source of continuity in the guidance of the affairs of the IMS is the Executive Committee. During this past year in the Executive Committee was the transition from Jean Opsomer to Zhengjun Zhang as IMS Treasurer. Jean served as IMS Treasurer for two full terms (2010–2016). His careful guidance of our financial affairs has put the IMS in a strong position for the foreseeable future. The other members of the Executive Committee are: Aurore Delaigle, who is currently serving the last year of two three-year terms as Executive Secretary; Judith Rousseau, also serving a second three-year term as Program Secretary (2012–2018); the President-Elect Alison Etheridge; the IMS Past President Richard Davis; and myself as IMS President.

**Journals**

The IMS has excelled in production of its flagship journals, the *Annals of Probability*, the *Annals of Statistics*, the *Annals of Applied Probability*, and the *Annals of Applied Statistics*. Producing these excellent journals continues to be one of the most important IMS activities. Our widely-read review journal, *Statistical Science*, has been in great editorial hands under the editorship of Peter Green who stepped down on January 1, 2017. Peter has been succeeded by Cun-Hui Zhang. In particular, the *Statistical Science* “conversation pieces” continue to be popular. Recently, in connection with the death of Peter Hall and the timely publication of a conversation article with Peter written by Aurore Delaigle and Matthew Wand, the IMS received several requests to make this conversation piece “open access”. In response to this request, the IMS Council recently decided to make all the conversation pieces completely open and accessible to everyone [*see this link*]: for a complete listing, from Hirotugu Akaike to Yuri Vasilyevich Prokhorov and Willem van Zwet, see: http://www.imstat.org/sts/conversations.html.

This past year was the tenth anniversary of the *Annals of Applied Statistics*, which has been a great success thanks to the excellent editorial work and leadership of past editors Brad Efron (2006–2012) and Steve Fienberg (2013–2015), and the current editor Tilmann Gneiting. The *Annals of Applied Statistics *has published about 2025 pages per year since it started in 2006. This year we also celebrate the tenth anniversary of one of our joint electronic journals, *Statistics Surveys*. This journal, a cooperative effort between the IMS, the Bernoulli Society, the Statistical Society of Canada and the American Statistical Association, was launched in 2007. *Statistics Surveys *has published on average 159 pages per year. In contrast, its sister journal *Probability Surveys *has published (on average) 322 pages per year. I believe that *Statistics Surveys *and *Probability Surveys *both have considerable potential for growth and increased importance, due to the communication role that well-written survey articles can play.

The *IMS Bulletin *Editor for the past three years has been Anirban DasGupta. Many thanks to Anirban for his dedicated editorial work and service to the IMS, and welcome to Vlada Limic who took over editorship of the *Bulletin *with this issue [*see her article*]. On behalf of all the members of the IMS, I would like to thank the editors of all the IMS journals and their editorial boards for all the superb work they do to maintain the high standards of our journals.

**IMS Meetings**

The last IMS Annual meeting was held jointly with the Bernoulli Society in Toronto during July 2016: the 9th World Congress in Probability and Statistics drew about 450 participants, and was a great success—in large part due to the excellent program organized by Alison Etheridge (Program Chair for the meeting and current IMS President-Elect), and groundwork by the local organizing committee chaired by Tom Salisbury. Our next IMS Annual meeting will be held during the Joint Statistical Meetings in Baltimore (July 29–August 3, 2017); while our next stand-alone meeting will be in Vilnius, Lithuania (July 2–6, 2018). Our next joint meeting with the Bernoulli Society will be the 10th World Congress in Probability and Statistics in Seoul, Korea (August 17–21, 2020).

**CIS wind-down and planned closure**

While our journals have been doing well, our long-time indexing, bibliographical, and retrieval effort, the *Current Index to Statistics*, has struggled with changing technologies and lack of use, with declining subscriptions and decreasing revenues. Most of the relevant scientific bibliography is now covered by Google Scholar, a transition presaged by the January/February 2005 *Bulletin* cover article about Google Scholar entitled “Stand on Giants’ Shoulders”. Although *CIS* has had a group of devoted users, anecdotal evidence suggests that many, if not most, IMS and ASA members are not more than occasional users—if they are even aware of its existence. The *CIS* Management Committee discussed various alternatives in person at the JSM in Chicago and online after the Chicago meeting. Following discussions by the IMS Executive Committee and the Council, the IMS Council recently voted to phase out *CIS* over a three-year period (2017–2019) by making the current *CIS* database “open access” (i.e. freely available) in its current form (with updates only from the automated processes currently in place and no further manual updates). The automatic updates will be dropped on January 1, 2020, and an effort will be made to share the current data with some other partner (to be determined). Announcements of these changes have been made in e-mails to subscribers, in the previous issue of the *Bulletin*, in the IMS e-bulletin, and on the *CIS* web pages.

While I regret the necessity of phasing out the *CIS* effort, I would like to convey sincere thanks to the current and past Editors (George Styan, Alan Zaslavsky, Eric Suess, Pantelis Vlachos), as well as others on the *CIS* Management Committee (David Umbach [chair], Chris Bilder, Xinping Cui and Haydar Demirhan), for their dedication and commitment to the CIS.

**IMS Groups**

One of the important projects of my predecessor, Richard Davis, has been a push to re-invigorate the IMS Groups program. One success in this direction has been the **New Researchers Group** (NRG), spearheaded by Alex Volfovsky, Dan Sussman, Vince Lysinski and others. The NRG has secured NSF Funding for the annual New Researchers Conference and has a new website thanks to Dan Sussman (http://groups.imstat.org/newresearchers/). The new website for the NRG has incorporated an on-line version of the IMS New Researchers’ Survival Guide, originally published in 2006 (see the January 2006 *IMS Bulletin*).

Another potential group in the direction of **Machine Learning** has been under discussion, but has not yet gotten off the ground. An ever-present need exists for the application of statistics and other science based methods to a wide range of problems, and this might be one avenue for the organization of several new IMS Groups, for example in the area of Environmental Statistics. If you have ideas or suggestions concerning potential new Groups in any area of probability or statistics, please contact us.

**Membership**

The IMS needs to continue to work to recruit new members. We had an organized membership drive in 2008 (see the *IMS Bulletin* Jan/Feb 2008), but it may be time for another drive. As noted by past Presidents Richard Davis (see the *IMS Bulletin *Jan/Feb 2016) and Erwin Bolthausen (in his IMS 2015 Presidential address, September 2015 *Bulletin*), the IMS has apparently lost some membership in the probability part of our community. Please consider persuading your colleagues and students, who are not already members, to join. Student membership is still free (see http://imstat.org/membership/student.htm), and membership for those doing work or teaching in probability or statistics is a bargain at $105 (recently reduced from $115, and with a further 10% reduction to $94.50 for early renewal before December 31) for a basic membership.

**Nominations**

The procedure for nominations has changed. In particular the IMS has opened the nomination process for named lectures and for proposing sessions at IMS sponsored meetings. I would like to encourage members to take advantage of these opportunities to provide direct input in both of these directions. For further information see announcements in the *IMS Bulletin *and e-bulletin.

**Invitation**

The IMS depends on its members for ideas and service in many different roles: editing our journals, organizing our meetings, and serving on committees to choose members for various honors and awards. I would like to thank our members who have volunteered their time and energy to serve in these important roles (with apologies for not being able to name you all here). I invite all members to step forward to serve the IMS.

In closing, let me re-iterate my invitation to all members to feel free to contact me (email to president@imstat.org), any member of the Executive Committee, or the IMS Executive Director Elyse Gustafson (email to erg@imstat.org) with your comments and suggestions.

Your ideas for improving the IMS and for new IMS initiatives are always welcome.

]]>With this issue I officially take the baton from Anirban DasGupta, to whom I am very grateful for all the good advice during my training in the previous months. We are working to complement the following list of Contributing Editors: Anirban DasGupta, David Hand, Xiao-Li Meng, Dimitris Politis and Terry Speed. Many thanks to them for staying on the team, to Anirban for bringing his “Angle” back and continuing with the Student Puzzle Corner (look out for these in the coming issues), as well as to outgoing Contributing Editors Robert Adler, Peter Bickel, Stéphane Boucheron and Hadley Wickham for their important service over the past years. Our hope is that the *Bulletin* continues its role in serving our community in the best possible way, without necessarily your noticing a transition taking place.

**Send us your POPI submissions**

Over the coming months we shall develop the “POPI” board—a space to report on a **Project, Object or Perspective of (potential community) Interest**. These will usually include important input from a peer who tells us about their favorite POPI of the moment. In case of a temporary lack of POPIs gathered from you, I may take one from my pool (or pick one from my garden?). Please send your POPI ideas to bulletin@imstat.org. Your input is vital for this initiative to succeed!