Deb Nolan writes:
The Institute of Mathematical Statistics has joined the INGenIOuS community (Investing in the Next Generation through Innovative and Outstanding Strategies), along with the American Statistical Association, US National Science Foundation, Mathematical Association of America, American Mathematical Society, and Society for Industrial and Applied Mathematicians. INGenIOuS will host a series of online and in-person events to develop strategies for investing in the training of the next generation of undergraduate and graduate students. The aim is to engage the mathematical sciences community in thinking strategically about enhancing recruitment, retention, and job placement in our community. The discussion has been divided into the following six sections (descriptions from the INGenIOuS website):

Recruitment & Retention – Helping to make the mathematical and statistical sciences a vibrant choice for a broad segment of the population (including the issue of broadening participation of women and minorities).

Technology & MOOCs – The expanding role of technology and its uses across STEM fields (the new opportunities it is bringing about in terms of new science; alternate forms of course delivery like MOOCs (massive open online courses) or flipped classrooms, how the introduction of new technology presents new challenges in terms of training).

Internships – Fostering and enhancing internships, co-ops, and industrial training opportunities for students at all levels.

Job Placement – Current best practices for connecting mathematical and statistical sciences students to jobs in all sectors

Measurement & Evaluation – Measures and mechanisms to assess the efficacy of, and return on investment in, the variety of successful training activities that departments are offering. How do we know we know that any particular program made a difference?

Documentation & Dissemination – The documentation and dissemination of effective training practices. How does the community avoid re-inventing the wheel without being too prescriptive?

The IMS would like to encourage its members to add their voice to these important discussions. Your participation is essential to the success of this effort. To be part of this project and to participate in the online discussions and panels please join the INGenIOuS community at www.ingeniousmathstat.org and consider participating in the following opportunities:

• Complete a survey at http://fs24.formsite.com/ingenious/form2/index.html
• Attend an online panel (if you read this in time): Internships (May 1), Job Placement (May 9), Measurement and Evaluation (May 17), Technology and MOOCs (May 30), Documentation & Dissemination (May 31), Recruitment & Retention (June 7).
• Joining a brief online open discussion following each panel.
• Apply to attend the final three-day workshop July 14–16.

For more information on each of these activities, visit http://www.ingeniousmathstat.org/

US National Academy of Sciences elects Peter Hall and Greg Lawler

The US National Academy of Sciences has elected 84 new members and 21 foreign associates from 14 countries in recognition of their distinguished and continuing achievements in original research. Among them are two familiar names: former IMS President Peter Hall, and IMS Fellow Gregory F. Lawler. Peter Hall is Australian Laureate Fellow in the Department of Mathematics and Statistics at the University of Melbourne, Australia, and Distinguished Professor at UC Davis. He was elected a Foreign Associate. Greg Lawler is professor in the Departments of Mathematics and Statistics at the University of Chicago. Members are elected to the National Academy of Sciences in recognition of their distinguished and continuing achievements in original research. National Academy membership is considered one of the highest American honors that a scientist can receive.

Larry Shepp, 1936–2013

On April 23 Larry Shepp, the Patrick T. Harker Professor in the Statistics Department at the Wharton School of the University of Pennsylvania, passed away at the age of 76, having been unable to recover from a fall several months ago. Larry was loved by many and had friends all over the world. Internationally recognized as a distinguished mathematician and probabilist of the highest caliber, Larry was an elected member of the National Academy of Sciences, the National Institute of Medicine, and the American Academy of Arts and Sciences. A full obituary will follow.

Gareth Roberts and Terry Speed named Fellows of the Royal Society

The UK’s Royal Society is a Fellowship of the world’s most eminent scientists and is the oldest scientific academy in continuous existence. Each year it elects new Fellows from the UK and Commonwealth, and Foreign Members; they are elected on the basis of excellence in science. There are approximately 1,450 Fellows and Foreign Members, including more than 80 Nobel Laureates. Among those elected this year are Gareth Roberts and Terry Speed. Gareth Roberts, University of Warwick, UK: according to http://royalsociety.org/people/gareth-roberts/ his work spans “applied probability, Bayesian statistics and computational statistics. He has made fundamental contributions to the theory, methodology and application of Markov Chain Monte Carlo and related methods in statistics. He has developed crucial convergence and stability theory, constructed a theory of optimal scaling for Metropolis-Hastings algorithms, and has introduced and explored the theory of adaptive MCMC algorithms. He has made pioneering contributions to infinite dimensional simulation problems and inference in stochastic processes.” Terry Speed is Senior Principal Research Scientist at the Walter and Eliza Hall Institute of Medical Research. The Royal Society website http://royalsociety.org/people/terence-speed/ says Terry, “is regarded internationally as the expert on the analysis of microarray data. This results partly from the sheer ingenuity of his work, and in part it is due to his commitment to working closely with biomedical scientists, enabling him to appreciate first-hand the biological challenges and the consequent requirements of new methodology … [He] has made seminal contributions to bioinformatics, statistical genetics, the analysis of designed experiments, graphical models and Bayes networks.”

American Academy of Arts and Sciences elects Larry Brown, Bin Yu

The American Academy of Arts and Sciences has elected Larry Brown and Bin Yu to its membership. Founded in 1780, the American Academy of Arts and Sciences is an independent policy research center that conducts multidisciplinary studies of complex and emerging problems. The Academy’s elected members are leaders in the academic disciplines, the arts, business, and public affairs.

Lawrence David Brown, University of Pennsylvania’s Wharton School, lists his research interests at http://www-stat.wharton.upenn.edu/~lbrown/ as, “statistical decision theory; statistical inference; nonparametric function estimation; foundations of statistics; sampling theory (census data); empirical queueing science.” Bin Yu, Department of Statistics, University of California, Berkeley, is currently IMS President-Elect. According to her department webpage, Bin is “currently working on statistical machine learning theory, methodologies, and algorithms for solving high-dimensional data problems. Current research topics of my group cover sparse modeling (e.g. Lasso), structured sparsity (e.g. hierarchical and group and graph path), analysis and methods for spectral clustering for undirected and directed graphs; and our data problems come from diverse interdisciplinary areas including remote sensing, neuroscience, document summarization, and social networks. My past research areas have also included empirical processes, Markov Chain Monte Carlo, signal processing, the minimum description length principle (MDL), and information theory.”

The complete list of new members is at http://www.amacad.org/news/classlist2013.pdf

Eyal Lubetzky receives Rollo Davidson Prize

The Rollo Davidson Trustees have announced the award of the 2013 Rollo Davidson Prize jointly to Eyal Lubetzky (Microsoft Research, Redmond) and Allan Sly (University of California, Berkeley) for their work on the dynamics of the Ising model, and especially their remarkable proof of the cut-off phenomenon.

New IMS Managing Editor

IMS Council has approved the appointment of T.N. Sriram as Managing Editor, for the term January 1, 2014 to December 31, 2016. He will take over from Michael Phelan. T.N. Sriram is a professor in the Department of Statistics at the University of Georgia, Athens. See his webpage at http://www.stat.uga.edu/people/faculty/tn-sriram

Vincenzo Capasso awarded “Chair of Excellence”

Vincenzo Capasso, who is a member of IMS and an Elected Fellow of ISI, is Full Professor of Probability and Mathematical Statistics at the Department of Mathematics, Milan University, Italy. He has been awarded one of ten Chairs of Excellence for the 2013–14 academic year, in an international competition called by Carlos III University of Madrid, in order to promote excellence in research and attract frontline researchers from the international university and research community. The awardees in all fields of research were selected by an evaluation committee composed of eight senior professors, including five from Carlos III.

COPSS Fisher Lecture by Peter Bickel

Peter Bickel will give the COPSS Fisher lecture at JSM Montreal on August 7th, at 4pm. The title of his talk is From Fisher to “Big Data”: continuities and discontinuities.

Nominations are sought for the 2013 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. The prize of $1,000 will be presented at the NISS/SAMSI JSM Reception on August 5, 2013, in Montreal.

Further information about the award can be found at www.niss.org/news/awards/jerome-sacks-award-outstanding-cross-disciplinary-research

To nominate an individual, please submit a nomination letter (maximum two pages, including the names of at least two other individuals who have consented to write letters of support) and a CV, to sacksaward2013@niss.org

Currently, the Journal of Statistical Software is the only FOAS project. JSS has grown rapidly over the 15 years of its existence, in page count, quality, and impact. The journal does not charge fees to authors or to readers. It needs a more stable support structure to guarantee its continued existence and growth.

On the FOAS website you can join, and/or make financial contributions. We invite you to contribute ideas, projects, and materials for the FOAS site.

Jan de Leeuw, email: jan.deleeuw@foastat.org
Katharine Mullen, email: katharine.mullen@foastat.org
Achim Zeileis, email: achim.zeileis@foastat.org

The IMS members’ responses have now been collated into a report, which is available for download here.

Donald Lyman Burkholder died in his sleep on April 14, 2013, in Urbana, Illinois. He was born January 19, 1927, in Octavia, Nebraska, the fourth of five children of Elmer and Susan (Rothrock) Burkholder. His mother had been a schoolteacher, and his father was a farmer who served on the community school board for many years. Education became the family business: of the four boys, the oldest was a superintendent of schools, the three youngest were college professors, and many in the next generation are educators.

In 1945, Don graduated from high school, where he was captain of the basketball team and senior class president, an honor (as he loved to relate) that came his way because his three classmates had already been president. He was drafted and entered the Civilian Public Service (CPS) as a conscientious objector, serving as a cook at a camp for fighting forest fires in Oregon and as an orderly at a mental hospital in New Jersey.

Following his discharge in December 1946, he acted on the recommendation of a friend and enrolled at Earlham College, a predominantly Quaker college in Richmond, Indiana. There he met his wife-to-be, Jean Annette Fox, and they were both drawn to the field of sociology by the vision and intellectual rigor of a new faculty member who had also served in the CPS, Bill Fuson.

After their wedding in June 1950, Don and Jean attended the University of Wisconsin in Madison as graduate students in sociology. In 1953, they went to the University of North Carolina at Chapel Hill, where Don had a fellowship to study sociological statistics. He soon discovered that his real interest lay in mathematics, and he completed a PhD in mathematical statistics in 1955 under the guidance of Professor Wassily Hoeffding. That summer, Don joined the Mathematics Department at the University of Illinois, Urbana-Champaign. In 1978, he was appointed professor in the Center for Advanced Study, allowing him to devote more time to research. He retired as professor emeritus in 1998.

Soon after he came to Illinois, Don, influenced by his eminent colleague Joseph Doob, turned to the study of martingales. It is now apparent that the two mathematicians who most advanced martingale theory in the last seventy years were Joseph Doob and Donald Burkholder. Martingales as a remarkably flexible tool are used throughout probability and its applications to other areas of mathematics. They are central to modern stochastic analysis. And martingales, which can be defined in terms of fair games, lie at the core of mathematical finance. Burkholder’s research profoundly advanced not only martingale theory but also, via martingale connections, harmonic and functional analysis.

In their 1970 Acta Mathematica paper, which followed Burkholder’s seminal 1966 paper “Martingale Transforms” in the Annals of Mathematical Statistics, Burkholder and Gundy introduced a remarkable technique which shows how certain integral inequalities between two nonnegative functions on a measure space follow from inequalities involving only parts of their distribution. This seemingly simple but incredibly elegant technique, now referred to simply as “the good–λ method”, revolutionized the way probabilists and analysts think of norm comparison problems. It is now widely used in areas of mathematics which involve integrals and operators. Burkholder’s outstanding work in the geometry of Banach spaces arose from his extension of martingale inequalities to settings beyond Hilbert spaces where the square function approach used in his earlier work fails. His work in the eighties and nineties on martingale inequalities with emphasis on identifying best constants has become of great importance in recent years in the investigations of two well known open problems, one concerning optimal $L^p$ bounds of certain singular integrals operators and their ramifications in quasiconformal mappings and the other related to a longstanding conjecture in the calculus of variations dealing with rank-one convex and quasiconvex functions. These problems come from fields which on the surface are far removed from martingales.

The paper of Burkholder and Gundy mentioned above and the 1971 Transactions of the American Mathematical Society paper of Burkholder, Gundy, and Silverstein, are exceptionally important. The first paper includes, in addition to the good–λ inequalities, fundamental integral inequalities comparing the maximal function and the square function of martingales. The second paper strikingly improved, and completed, work of Hardy and Littlewood on the characterization of the Hardy $H^p$ spaces via the integrability of certain maximal functions. While probabilistic techniques had already gained the respect of many analysts studying harmonic functions and potential theory, due to earlier work of Doob, Kakutani, Wiener and others, this landmark paper had a profound influence in harmonic analysis and propelled many analysts to learn probability.

In his five-decade career, Don gave hundreds of invited lectures and lecture series at universities all over the world. He lectured in England, France, Germany, Switzerland, Israel, Denmark, Sweden, Poland, Hungary, Japan, Singapore, Italy, Scotland, Spain, and Canada and at universities across the United States. He was editor of the Annals of Mathematical Statistics (1964–67), president of the Institute of Mathematical Statistics (1975–76), and a member of many councils, advisory committees, and governing boards. He was a dedicated teacher and mentored 19 PhD students. He was elected to the National Academy of Sciences in 1992, and was a Fellow of the American Academy of Arts and Sciences, the Society for Industrial and Applied Mathematics, and the American Association for the Advancement of Science. In December 2012, he was among the first class named as Fellows of the American Mathematical Society.

Don will be remembered not only for his profound contributions to mathematics, but also for the kind and decent ways in which he interacted with everyone he met, and for his encouragement and support to so many young mathematicians who had the great fortune of crossing paths with him.

Don was predeceased by his brothers Robert and Wendell Burkholder and his daughter Kathleen Linda Burkholder; and is survived by his wife of almost 63 years, Jean Annette (Fox) Burkholder; his son J. Peter Burkholder and son-in-law P. Douglas McKinney of Bloomington, Indiana; his son William F. Burkholder, daughter-in-law Joanne (McLean) Burkholder, and granddaughter Sylvie Kathleen Burkholder of Singapore; his sister Helen Dale and brother-in-law Ernie Dale of Auburn, Washington; his brother and sister-in-law John and Donna Burkholder of McPherson, Kansas; his sisters-in-law Anne Burkholder of McPherson, Kansas, and Leona Burkholder of Madison, Wisconsin, and 17 nieces and nephews

Written by Peter Burkholder, Indiana University; William Burkholder, Institute of Molecular and Cell Biology, Singapore; Rodrigo Bañuelos, Purdue University; Burgess Davis, Purdue University; and Renming Song, University of Illinois at Urbana-Champaign

Martin B. Wilk, OC, died in Yorba Linda, California, on February 19, 2013; he was 90.

Throughout his career, Martin demonstrated that a statistician can successfully span academia, industry and government. For over half a century, he made important contributions, and occupied senior positions, in each of these domains. While his name may be best known within the profession for the Shapiro–Wilk test for normality, his influence on statistical methods and practice has been much broader. He was, among others, Assistant Vice President and Director of Corporate Planning at AT&T and Chief Statistician of Canada. In 1999, he was made an Officer of the Order of Canada (OC) for providing “insightful guidance on important matters related to our country’s national statistical system.”

Born and raised in Montréal, Martin Wilk attended McGill University, where he completed his Bachelor’s degree in Chemical Engineering in 1945. After graduation he joined Canada’s National Research Council atomic energy project at Chalk River, Ontario, where he soon recognized the critical role of variability in data analysis. At first he developed his own techniques to handle this variability. It was only after his move in 1950 to Iowa State College as a Laboratory Research Assistant that he discovered the discipline of statistics. He was soon enrolled in statistical courses and underwent his conversion from engineer to statistician. At Iowa he completed a Master’s degree in 1953 and a PhD in 1955 in the area of experimental design under the supervision of Oscar Kempthorne.

Martin’s postdoctoral year was spent at Princeton University under John Tukey during which he was introduced to the research work of Bell Labs on a part-time basis. Attracted by the research environment of Bell Labs, he chose to continue there after his postdoctoral year. During the 1960s he took on progressively more senior positions in the statistical methods and research groups of Bell Labs. Between 1959 and 1963 he was also Professor of Statistics at Rutgers University, New Jersey, while maintaining a part-time consulting relationship with Bell Labs.

The contributions to statistical methodology for which Martin Wilk is renowned stem largely from this period at Bell Labs. With primary collaborators, Ram Gnanadesikan and Samuel Shapiro, he published a series of papers dealing with probability plotting for multivariate data, and diagnostic procedures for classical distributions, including the well-known Shapiro–Wilk test statistic for normality.

By the end of the 1960s, Martin had developed an interest in the broader managerial and organizational issues of the American Telephone and Telegraph Company (AT&T), the parent company of Bell Labs. Beginning with the issue of rate setting for telephone services, and as preparation for hearings by the Federal Communications Commission, he became involved in the assessment and improvement of the models being used to value various business lines. This involvement led to a recognition by AT&T’s management of the broader value of management science and Martin was to be an in-house leader in this respect. During the 1970s he directed staffs involved in corporate modeling, research and planning, becoming Assistant Vice President and Director of Corporate Planning in 1976.

In 1980, Martin was approached by the Government of Canada for the position of head of Statistics Canada. At that time, the agency had been experiencing some serious difficulties. Independent reviews of both management and of methods, commissioned by the Government, had identified a range of issues that needed to be addressed, not least of which was a loss of staff morale following a period of adverse publicity. Martin accepted this challenge and became Chief Statistician of Canada late in 1980, the first mathematical statistician to occupy this post.

Between 1981 and 1985, Martin refocused Statistics Canada by, for example, introducing a more integrated and cohesive organizational structure, strengthening the Agency’s contacts with Ministries and other important data users, putting in place a disciplined planning system, rationalizing its program of publications, and establishing a stronger analytical capacity. He gave the organization a sense of purpose again. During this period he also had to deal with a sudden Cabinet decision to cancel the 1986 census, a decision that he managed to have reversed after some persuasive lobbying and innovative funding proposals. Martin’s short tenure as Chief Statistician of Canada set the stage for Statistics Canada to flourish and become recognized as a world-class statistical agency over the following two decades.

After his retirement from Statistics Canada in 1985, Martin remained in Ottawa and undertook several important consultancies for the Canadian Government. In particular, he headed the National Task Force on Health Information that led to the creation of the Canadian Institute of Health Information. He also conducted a review for Revenue Canada of their data management and holdings with emphasis on strengthening the statistical use of these data. He served for many years on the National Statistics Council of Canada as well as on Statistics Canada’s Advisory Committee on Statistical Methods. Finally, approaching 80, he retired to the West Coast of the United States, where he was able to enjoy his later years with his second wife, Dorothy, his children and grandchildren.

Martin was a Vice President of the American Statistical Association in 1980–82, having previously served as President of his local chapter. He was also President of the Statistical Society of Canada for 1986–87, promoting the strengthening of ties between academic statisticians and statisticians in industry and government. He was made an Honorary Member of the SSC in 1988 “for seminal contributions to the fields of analysis of variance, multivariate analysis, model fitting and validation, for enormous contributions to Statistics Canada as the Chief Statistician and for insightful guidance of the Society while serving on its Board and as its President.”

Martin Wilk’s contributions to the statistical profession were recognized by many other honors throughout his career. He was, among others, a fellow of the ASA (1962), the IMS (1968), and the American Association for the Advancement of Science (1969). He received the Jack Youden Prize in 1972 and a Distinguished Alumni Award from Iowa State University in 1997.

Those who worked with Martin recall his formidable ability to argue a case, extemporaneously and sometimes at length, his penetrating questions often from unexpected angles, his ability to analyze complex issues quickly and focus on the crux of the matter, and his unending supply of aphorisms exactly suitable for the issue at hand. Many of his pronouncements continued to be quoted at Statistics Canada long after he retired.

The profession has lost a great statistician whose contributions to theory and practice will long be influential. For additional information about Martin Wilk’s life and career, see, among others:

D.R. Brillinger & J.F. Gentleman (1989). Martin Bradbury Wilk: a new honorary member of the Statistical Society of Canada. Liaison, 4 (1), 27–29.

C. Genest & G.J. Brackstone (2010). A conversation with Martin Bradbury Wilk. Statistical Science, 25, 258–272.

M. Lennick (2013). Martin Wilk remembered as ‘the best statistician in Canada’s history’. The Globe and Mail, Toronto, April 13.

Written by Christian Genest, McGill University, and Gordon J. Brackstone, Statistics Canada

A Priori Analysis of Complex Models

We (P.J. Bickel, A.C. Gamst, B.J.K. Kleijn, and Y. Ritov) study a few examples of Bayesian procedures on complex, high-dimensional parameter spaces. The Bayesian procedures we consider are those that adhere to the following paradigm. The prior distribution is announced prior to observing the data. At least we are restricted to priors that do not depend on details of the experimental design or on knowing the specific functions of the parameters that may turn out to be of interest. In this paradigm, it would not, for example, be reasonable for a statistician to use one prior for estimating one function, and another to estimate a different function. We shouldn’t be reminded of Groucho Marx’s quote, “Those are my principles, and if you don’t like them… well, I have others.”

Bayesian procedures can be considered from different points of view. Their closure is the set of admissible procedures, and they are known to generate asymptotic minimax procedures in regular parametric models. However, these claims are valid when the priors are selected to fit frequentist ad-hoc considerations.

Most early discussions of Bayesian analysis presented simple examples, e.g., X ~ N(ϑ, 1). In this case, a statistician might have clear a priori ideas about ϑ, and might well understand the implications of using his prior. Regardless, the data will eventually overwhelm the prior, and typically frequentist and Bayesian inference will coincide. The classical Bernstein–von Mises Theorem encapsulates this observation. Currently, Bayesian procedures are being applied to complex, high-dimensional models, e.g., those used in medical imaging. With a very high-dimensional parameter space (where, for example, laws of large numbers appear in surprising places), it is very difficult to understand the implications of using a particular prior. It is very difficult, if not impossible, to express subjective information about the model in a robust prior, and it is difficult to express this knowledge in a way that would support the data analysis and not dominate it.

We use several examples to illustrate a number of issues. This includes the partial linear model of Engle, Granger, Rice and Weiss (1986) , and different models in the very convenient lab of white noise series. We show that in situations where the nonparametric part of the model is smooth enough, the Bernstein–von Mises phenomenon holds and Bayesian estimators are efficient, but the Bayesian estimator is going to fail in extreme situations where simple frequentist estimation can still work. Then, it may argue that in a given white noise model, the any Bayesian prior would fail in estimation of some linear functional, while trivial frequentist estimator would not.

We also give an example in which Bayesian procedures which ignore the stopping time associated with the data generating process fail, while simple frequentist procedures continue to work. This demonstrates the danger of the classical principle that Bayesians need not pay attention to stopping times.

The Mathematics of Causal Inference

Recent developments in graphical models and the logic of counterfactuals have had a marked effect on the way scientists treat problems involving cause–effect relationships. Paradoxes and controversies have been resolved, slippery concepts have been demystified, and practical problems requiring causal information, which long were regarded as either metaphysical or unmanageable, can now be solved using elementary mathematics.

I will review concepts, principles, and mathematical tools that were found useful in this transformation, and will demonstrate their applications in several data-intensive sciences. These include questions of confounding control, policy analysis, misspecification tests, mediation, heterogeneity, selection bias, missing data and the integration of data from diverse studies.

These advances owe their development to two methodological principles. First, a commitment to understanding what reality should be like for a statistical routine to succeed and, second, a commitment to express the understanding of reality in terms of data-generating models, rather than distributions of observed variables.

Data generation models, encoded as nonparametric structural equations, have led to a fruitful symbiosis between graphs and counterfactuals that has unified the potential outcome framework of Neyman, Rubin and Robins. with the econometric tradition of Haavelmo, Marschak and Heckman.

In this symbiosis, counterfactuals emerge as natural byproducts of structural equations and serve to formally articulate research questions of interest. Graphical models, on the other hand, are used to encode scientific assumptions in a qualitative (i.e., nonparametric) language, identify their testable implications, and determine the estimability of interventional and counterfactual research questions.

One of the major results along this development has been a complete solution to the problem of non-parametric causal effects identification. Given data from observational studies and qualitative assumptions of how variables relate to each other causally, it is now possible to decide algorithmically whether the assumptions are sufficient for identifying causal effects of interest, what covariates should be measured (or enter into a propensity score routine) and what the testable implications are of the model assumptions. “Completeness” proofs that accompany these results further assure investigators that no method can do better without resorting to stronger assumptions.

Another triumph of the symbiotic analysis has been the emergence of active research in nonparametric mediation problems, aiming to estimate the extent to which an effect is mediated by various pathways or mechanisms (e.g., Robins and Greenland, Pearl, Petersen and Van der Laan, VanderWeele, Imai). The importance of this analysis, aside from telling us “how nature works,” lies in policy evaluation, especially in deciding what nuances of a given policy are likely to be most effective. Mediation-related questions were asked decades ago by Fisher and Cochran but, lacking the tools of graphs and counterfactuals they could not be addressed until quite recently.

Recent works further show that causal analysis is necessary in applications previously thought to be the sole province of statistical estimation. Two such applications are meta-analysis and missing data.

The talk will focus on the following questions:

1. What every student should know about causal inference, and why it is not taught in Statistics 101. http://ftp.cs.ucla.edu/pub/stat_ser/r350.pdf

2. The Mediation Formula, and what it tells us about “How nature works” http://ftp.cs.ucla.edu/pub/stat_ser/r379.pdf

3. What mathematics can tell us about “external validity” or “generalizing across populations” http://ftp.cs.ucla.edu/pub/stat_ser/r372.pdf, http://ftp.cs.ucla.edu/pub/stat_ser/r387.pdf

4. When and how can sample-selection bias be circumvented http://ftp.cs.ucla.edu/pub/stat_ser/r381.pdf, http://ftp.cs.ucla.edu/pub/stat_ser/r405.pdf

5. What population data can tell us about unsuspected heterogeneity http://ftp.cs.ucla.edu/pub/stat_ser/r406.pdf

6. Why missing data is a causal problem, when parameters are estimable from partially observed data, and how http://ftp.cs.ucla.edu/pub/stat_ser/r406.pdf

Reference: J. Pearl, Causality (Cambridge University Press, 2000, 2009) Working papers: http://bayes.cs.ucla.edu/csl_papers.html