Info-Metrics Institute College of Arts and Sciences, American University Fall 2011 Newsletter Director’s Message Info-metrics is the science and art of processing information. Generally speaking, all information is limited and all data are noisy. In practice, this means that all inferential problems are under-determined as all scientific theories and models are based on limited and, at times, very scarce information. A major challenge for all scientists is how to reason under conditions of incomplete information. The Info-Metrics Institute and its affiliates work to address that challenge. Contents Director’s Message..........................1 Upcoming Events.............................1 Sponsor Spotlight...............................2 Grad Students In Their Words...........2 Notes from the Advisory Board............3 New Research Associates...................3 Our Work.........................................4 Affiliate Profile: Raphael Levine......7 Affiliate Profile: Essie Maasoumi....7 Thank You, ARO! ............................7 Short Term Visiting Fellows.............8 Institute Seminars.............................8 Recent Events..................................8 In the Institute’s second year, we expanded our activities and focused on building the interdisciplinary aspects of the program. We welcomed visiting fellows and researchers from across the scientific spectrum and hosted two major meetings – the first with an emphasis on information-theoretic Roundtable participants (L-R: Alistair Hall, Duncan Foley, Ariel Caticha, Aman econometrics, statistics and Ullah, Yuichi Kitamura, Eric Renault, Robin Lumsdaine, Peter Robinson and Michael Stutzer) at the September 2010 conference modeling; the second focus- ing on the fundamental questions of inference and modeling within the natural sciences. A major objective of both sessions was research into the foundations of info-metrics and entropic inference. Our proceedings and special issue journal publication summarize much of this new research. In 2010-11, the Institute’s successes included: • an increase in participants of the Institute’s core interdisciplinary research group; • one major conference and one workshop (September 2010 and May 2011); • nine short term fellows (between 3 to 14 days each); • an affiliates’ meeting following each conference to discuss activities and research (about 25 affiliates); • two graduate student summer fellowships; • two short, hands-on classes taught by Ingmar Prucha and Amos Golan; American University Info-Metrics Institute Department of Economics Kreeger Hall 4400 Massachusetts Ave, NW Washington, D.C. 20016-8029 202-885-3770 202-885-3970 (fax) info-metrics@american.edu www.american.edu/info-metrics • a working paper series and an updated online resource page. We would like to thank the Office of the Comptroller of the Currency for its continued support as well as the Army Research Office for its generous one-time grant. Their support helped fund the majority of the Institute’s activities. We plan to keep pushing the research frontiers of info-metrics to develop a leading interdisciplinary Institute. Bookmark our Web site and check in frequently for updates. And please, give us your feedback and ideas – we welcome your suggestions. - Amos Golan, Director, Info-Metrics Institute Upcoming Institute Events October 2011 Workshop: Philosophy of Information (October 3, 2011) November 2011 Workshop: Information Theory and Shrinkage Estimation (November 12, 2011) Spring 2012 Workshop: Information and Econometrics of Networks (March 30-31, 2012) All events will be held at American University. For more information, please visit our website. Info-Metrics Institute Fall 2011 Newsletter 2 The Info-Metrics Institute and Financial Reform The Office of the Comptroller of the Currency (OCC) supervises and regulates many of the largest banks in the country, banks that use increasingly complex modeling techniques in a continuously expanding range of applications. The OCC is pleased to have supported the Institute since its creation, in part because the Institute provides a way for OCC staff to enhance their modeling expertise through ongoing exposure to an active research community. OCC staff have adopted methods based in information theory for their own research; in addition, the activities of the Institute have helped make OCC staff better able to evaluate methods used by banks and verify that models are not used in ways that threaten the health of individual banks or the banking system. The recent financial crisis highlights another potentially important contribution of the Institute to the broader mission of the OCC. Analysts have cited various factors as potential contributors to the crisis – among them the failure of at least some of the quantitative models used in the financial sector. Although models were not the sole or even a primary cause of the crisis, some problems did stem from the use of analytical methods that were less effective than they should have been. The financial crisis emphasized that shortcomings in modeling approaches can contribute to significant losses for investors and financial firms, and impose serious economic costs on society more broadly. for control of “model risk,” defined as the potential for adverse consequences from decisions based on incorrect or misused model outputs. One of the themes of this new guidance is that effective use of models by banks requires processes for ongoing improvement, including exploration and evaluation of new and better techniques and data. This makes it incumbent on modeling experts to pursue new and better methods and approaches for use in financial settings. The OCC and other regulators around the world continue working to strengthen the financial system through various initiatives as part of broader financial reform. But it is clear that improvements in financial modeling can play an important role in making the financial system less susceptible to failure. The Institute is well positioned to contribute to methodological improvements that can make a practical difference in the operation of banks and other financial firms. In this respect, the Institute supports reforms that ultimately strengthen the global financial system and help avert future crises. The OCC is pleased to see the development and growth of an organization like the Info-Metrics Institute with the potential for such broad and substantial benefits. - Mark Levonian, Senior Deputy Comptroller, Office of the Comptroller of the Currency Drawing in part from the lessons of the financial crisis, the OCC recently published guidance for banks clarifying the OCC’s expectations Info-Metrics September 2010 workshop participants gather before the beginning of a session Graduate Students: In Their Own Words “Over the summer, I studied the information theoretic basis of multivariate measures of well-being. The aim of this research is to gain an understanding of the changes in the distribution of non-monetary measures of well-being across countries over time. The Institute’s summer graduate fellowship provided me with a great opportunity to explore the literature and extend my understanding of these information theoretic measures, while further honing my research skills under the tutelage of Amos.” “As my current research involves the empirical examination of structural transformation in Nicaraguan agriculture, I am primarily seeking to make a probabilistic statement about entry/ exit, expansion, and contraction for different sizes/types of producers in Nicaragua’s agricultural sector. Such questions are typically approached with the use of Markov chain analysis, which is a method that is particularly suited to estimation via maximum entropy. Given that maximum entropy econometrics was unfamiliar territory until recently, I found the summer graduate fellowship to be extremely beneficial to my research in particular as well as my understanding of info-metrics in general.” - Skipper Seabold – Heath Henderson www.american.edu/info-metrics 3 Notes from the Advisory Board The Info-Metrics Institute gained momentum last year with two major conferences, the development of web resources, access to software, a working paper series, expansion of the active visitor associate program, and summer courses on info-metrics and spatial econometrics. The Institute’s program for the coming year is the most ambitious yet, with three conferences (including a spring 2012 workshop on info-metrics and social networks), more visiting associates, and initiatives to make info-metrics techniques and research more accessible. We thank the Institute’s funders, conference planners, speakers and participants for working with us and we look forward to continuing to build a special network of researchers and learners. - Robert Lerman, Chair We are pleased to have welcomed so many of you to the Institute this year through our conferences, workshops, and shortterm visitor program. The spirit of collegiality and interdisciplinary interest that we all embrace is what makes the Institute so special. Welcome, New Research Associates! ROSSELLA BERNARDINI PAPALIA Department of Statistical Sciences University of Bologna, Italy MARINE CARRASCO Département de Sciences Economiques Université de Montréal, Montréal, Canada A. BOWEN GARRETT McKinsey & Company, Center for U.S. Health System Reform Urban Institute, Health Policy Center NATHAN L. HARSHMAN Department of Physics American University ATSUSHI INOUE Department of Agricultural and Resource Economics North Carolina State University ILYA NEMENMAN Departments of Physics and Biology Population Biology, Ecology, and Evolution Graduate Program Neuroscience Graduate Program Computational and Life Sciences Initiative Emory University TAISUKE OTSU Cowles Foundation & Department of Economics Yale University SHERMAN ROBINSON Environment and Production Technology Division Development Strategies and Governance Division International Food Policy Research Institute Washington, D.C. VICTOR M. YAKOVENKO Department of Physics University of Maryland, College Park For those of us who have come to info-metrics from the social sciences, the spring 2011 workshop, “Info-Metrics Across the Sciences,” exposed us to a new area of scholarly inquiry and emphasized the extent to which entropic inference crosses disciplinary boundaries. I’m confident that our fall workshops on the Philosophy of Information and Shrinkage Estimation will similarly expand our understanding and create further discussion synergies. As always, the Advisory Board welcomes your suggestions. - Robin Lumsdaine, Member Help us design a new logo! We are seeking talented students, graphic designers, and artists to enter the Info-Metrics Institute logo design contest. The winning logo will be selected by Institute staff and featured on our website and all supporting materials. To enter, contact Aisha Malik at malik@american.edu. Info-Metrics Institute Fall 2011 Newsletter 4 Our Work: Promoting Interdisciplinary Research Following is a list of representative research and recent publications by some of the Institute’s affiliates. More detailed information can be found in the 2010-11 Info-Metrics Annual Report, which will be available on our website in November 2011. Pieter Adriaans (University of Amsterdam) Current research: §Philosophy of information, generalized Kolmogorov complexity (with Golan), mathematical basis for a theory of facticity, study of data sets that show so-called saw tooth curve dependencies (with Schultes) Recent publications: §Philosophy of information, Stanford Encyclopedia of Philosophy. Rossella Bernardini Papalia (University of Bologna) Current research: §Exploration of entropy-based methods in ecological inference with spatial dependence; small area estimation problems; and spatial econometric models for panel data James Bono (American University) Current research: §Investigation of entropic inference as a solution concept in noncooperative games (with Golan) §Development of a statistical strategic model of air carrier behavior §Development of statistical strategic models of cyber security for smart energy grids Mehmet Caner (NC State University) Current research: §Adaptive lasso with heteroskedastic data §Instrumental variable selection Recent publications: §A Near Minimax Risk Bound: Adaptive Lasso with Heteroskedastic Data in Instrumental Variable Selection (with M. Fan) Marine Carrasco (University of Montreal) Current research: §Generalization of empirical likelihood estimation to handle a continuum or a countable infinite number of moment conditions §Regularization techniques of the covariance matrix and its application to portfolio selection in finance Ariel Caticha (SUNY – Albany) Current research: §Development of general methods of entropic and Bayesian inference §Application of principles of entropic inference to the foundations of physics §Quantum theory and the derivation of the Schroedinger equation, of uncertainty relations, and an analysis of the quantum measurement problem §Application of principles of entropic inference to economic modeling by focusing attention on simple models of highly idealized economies (with Golan) Duncan Foley (New School for Social Research and Santa Fe Institute) Recent publications: §Global warming and economic externalities, Economic Theory (with Rezai and Taylor). §Model description length priors in the urn problem, Computable, Constructive and Behavioural Economic Dynamics, Routledge. §The long-period method and Marx’s theory of value, The Evolution of Economic Theory: Essays in Honour of Bertram Schefold, Routledge, Abington and New York. §Mathematical formalism and political-economic content, INET. §Dynamics of output and employment in the U.S. economy, NSSR SCEPA PERI Working Paper Series (with Basu). A. Bowen Garrett (McKinsey & Co. and Urban Institute) Current research: §Economic and applied econometric modeling of health care and policy §Using info-metric methods in micro-simulation modeling of health reforms Ramo Gencay (Simon Fraser University) Recent publications: §Trading frequency and volatility clustering, Journal of Banking and Finance (with Xue). §Financial Applications of Non-extensive Entropy, IEEE Signal Processing Magazine (with Gradojevic). Info-Metrics May 2011 workshop participants continues on page 5 www.american.edu/info-metrics 5 §Exploring instabilities and misspecification and their relationship with information theory §Errors-in-variables estimation with wavelets, Journal of Statistical Computation and Simulation (with Gradojevic). George Judge (University of California - Berkeley) §Investment horizon effect on asset allocation between value and growth strategies, Economic Modelling (with In and Kim). Recent publications: §An Information Theoretic Approach to Econometrics. Cambridge University Press (with Mittelhammer). §A family of empirical likelihood functions and estimators for the binary response model, Journal of Econometrics (with Mittelhammer). John Geweke (U. of Technology Sydney) Current research: §Pooling financial and macroeconomic models for improved prediction Ramo Gencay presents at the September 2010 conference Recent publications: §Optimal Prediction Pools, Journal of Econometrics. §Massively Parallel Sequential Importance Sampling with Markov Chain Renewal (with Amisano). Other updates: §Appointed Theil Chair in Econometrics at Erasmus University Rotterdam, succeeding Clive Granger. Amos Golan (American University) Current research: §Foundations of info-metrics and entropic inference §Info-metrics and extreme events; info-metrics modeling and applications Alastair Hall (University of Manchester) Current research: §Estimation of the parameters of economic and statistical models based on the information in moment conditions §Generalized Method of Moments and Generalized Empirical Likelihood, including moment selection and structural stability testing. §Applications of GMM and GEL to financial, macroeconomic and health economic data Nathan Harshman (American University) Current research: §Investigating how information, correlation, and entanglement can be used to understand interacting quantum systems and the emergent distinctions between the quantum and the classical world §Applying symmetry methods to characterize invariant measures of information and entanglement in physical systems Recent publications: §Observables can be tailored to change the entanglement of any pure state, Physical Review A. Atsushi Inoue (NC State University) Current research: §Macroeconometrics, time series econometrics, forecasting §Large deviations theory and econometric information recovery, Handbook of Empirical Economics and Finance (with Grendar). Nicholas Kiefer (Cornell University) Current research: §Bayesian methods in risk management and compensation design in banks and financial institutions §Incentive compatible elicitation and representation of information Recent publications: §The Bayesian Approach to Default Risk: A Guide, Rethinking Risk Measurement and Reporting: Vol II (with Jacobs). Robin Lumsdaine (American University) Current research: §Exploring the relationship between financial market perceptions and reality, and the role of news and information in shaping those perceptions §Testing conventional wisdoms in the inflation-linked bond market §How survey design affects participant responses and subsequent inference §The impact of the changing demographic landscape on the global financial markets Ilya Nemenman (Emory University) Current research: §Information channel capacities of molecular networks Recent publications: §Information channel capacities of molecular networks §Information theory and adaptation, Quantitative Biology, (Ed. M.E. Wall) Taisuki Otsu (Yale University) Current research: §Application of information theoretic method for Bayesian semiparametric analysis, such as moment condition models and single index models continues on page 6 Info-Metrics Institute Fall 2011 Newsletter 6 §Robustness analysis of information theoretic bootstrap inference by using breakdown point theory §Large deviation analysis of information theoretic approach for statistical estimation, hypothesis testing, and model selection techniques §Causal inference §The relationship between coherence, confirmation and causation §Applications of causal model search applied to genetic regulatory network search Jeffrey Racine (McMaster University) Richard Smith (University of Cambridge) Current research: §Regression spline methods in categorical predictors setup Current research: §Exploring the theory of estimation and inferential methods for econometric models specified through moment conditions §Kernel smoothing and regression splines Recent publications: §Additive Regression Splines with Irrelevant Categorical and Continuous Regressors, The Comprehensive R Archive Network (www.r-project.org), (with Ma). §Spline Regression in the Presence of Categorical Predictors, The Comprehensive R Archive Network (www.r-project. org), (with Ma and Yang). §The ‘crs’ package for R, The Comprehensive R Archive Network (www.r-project.org). Eric Renault (Brown University) Current research: §Variance targeting for heavy tailed time seriess §Inference with implied probabilities §Testing strength of identification in moment conditions models §Stochastic Volatility models for Option Pricing §Volatility Measurement with High Frequency Data in Finance Recent publications: §Nonparametric Instrumental Regression (with Darolles, Fan and Florens), Econometrica. §Efficient Derivative Pricing by Extended Method of Moments (with Gourieroux and Gagliardini), Econometrica. Other updates: §Joined Brown University as the C.V. Starr Professor of Economics, effective July 1, 2011. §Development of valid inferential methods when confronted by instruments weakly correlated with endogenous variables §Development of reliable inferential methods in the moment condition context based on the bootstrap, a computationally intensive tool that has had success in other applications Michael Stutzer (University of Colorado – Boulder) Current research: §Exploration of info-metric optimal hedging using futures contracts David Wolpert (NASA Ames Research Center) Current research: §Self-dissimilarity and complexity, quantification of selfdissimilarity for real world data (with Macready) §Predictive game theory, Bayesian approach to formulating distribution-valued concepts, entropic priors (with Bono) Recent publications: §Decision Making with Imperfect Decision Makers, Springer. Ximing Wu (Texas A&M University) Current research: §Nonparametric and information-theoretic methods in econometrics §Data-driven information-theoretic methods of distributional and specification hypotheses §Nonparametric estimation of multivariate density and copula density functions Sherman Robinson (International Food Policy Research Institute) §Semiparametric estimation of shape-constrained functions Current research: §Maximum entropy estimation methods, including estimating national accounts and Social Accounting Matrices (SAMs) for developing countries, and cross-entropy estimation of GIS data at the pixel level for countries in sub-Saharan Africa §Multi-attribute welfare inferences Richard Scheines (Carnegie Mellon University) Current research: §Psychometric models and methods Victor Yakovenko (University of Maryland) Current research: §Application of methods of statistical physics to economic, financial, and social problems §Exploration of probability distributions of money, wealth, income, and energy consumption per capita www.american.edu/info-metrics 7 Affiliate Profile: Esfandiar (Essie) Maasoumi Affiliate Profile: Raphael Levine Essie Maasoumi, Info-Metrics Institute Affiliate Raphael Levine, Info-Metrics Institute Affiliate Esfandiar “Essie” Maasoumi advances the objectives of the Institute by sharing his extensive knowledge in the disciplines of economics, econometrics, financial econometrics, information theory and information-theoretic estimation methods. Chemical Physicist Raphael “Raphy” Levine brings the interdisciplinary work of the Institute to life through his expertise in the foundation of information theory, maximum entropy and the fundamentals of the natural sciences. Essie’s wide-ranging areas of research include nonlinear time series, nonparametric estimation methods, tests for heteroskedasticity and structural change, measures of welfare and mobility, and the study of neural networks. Raphy is widely known for his contributions to the role of energy in chemical reactions, but his work also involves examining more abstract issues. “For this, methodologies based on information theory and on algebraic quantum mechanics are useful, as well as models that emphasize key aspects of the problem and allow for a simple conceptual picture of the dynamics as much as exact numerical simulations,” he says. In our new section, Affiliate Profiles, we share the stories of the people behind the Institute’s success and shed light on the interdisciplinary nature of their work. Our first installment features Essie Maasoumi and Raphy Levine. He serves as the Arts and Sciences Distinguished Professor of Economics at Emory University in Atlanta and he is the editor of Econometric Reviews. He has been a Fellow of the Journal of Econometrics since 1989 and a Fellow of the Royal Society since 1991. In 2011, he was named as a Fellow of the American Statistical Association. Essie’s 1993 compendium in Econometric Reviews is the first survey article on information theory in economics and econometrics. He is ranked 37th in the world in The Econometricians’ Hall of Fame. Thank You, ARO! The U.S. Army Research Laboratory’s Army Research Office (ARO) has co-sponsored two Info-Metrics Institute workshops to date. We wish to thank Dr. Janet Spoonamore and the Decision and Neurosciences Program of ARO for their generous support. Raphy’s work combines the principle of maximum entropy with information theoretic inversion procedures to the basics of chemical, physical and biological applied and theoretical research. He is a member of Academia Europaea, American Academy of Arts and Sciences, American Philosophical Society, International Academy of Quantum Molecular Science, Israel Academy of Sciences & Humanities, Max Planck Society, Royal Danish Academy of Sciences and Letters, and the U.S. National Academy of Sciences. He was Alfred P. Sloan Fellow and has honorary degrees from the Technische Universität München and from Liege University. He has been the Miller Research Professor at UC Berkeley, the A.D. While Professor at Large at Cornell, and the Max Born Professor of Natural Philosophy at the Hebrew University. He has received the Wolf Prize in Chemistry, the Max Planck Prize for International Cooperation and other awards. Raphy currently serves as distinguished professor of chemistry in the College of Letters and Science and as distinguished professor of medical and molecular pharmacology at the School of Medicine of UCLA. He also serves as a member of the editorial board of the Proceedings of the National Academy of Sciences and of other scientific publications. Info-Metrics Institute Fall 2011 Newsletter 8 Fall 2010/Spring 2011 Institute Visiting Fellows Ramo Gencay (Simon Fraser University) Luca Secondi (University of Florence) Alastair Hall (University of Manchester) Ximing Wu (Texas A&M University) Eric Renault (Brown University) Aman Ullah (University of California, Riverside) Steve Pincus Pieter Adriaans (University of Amsterdam) Ariel Caticha (SUNY–Albany) Recent Institute Events FALL 2010 WORKSHOP INFO-METRICS: THEORY AND APPLICATIONS IN THE SOCIAL SCIENCES (September 24-25, 2010) SPRING 2011 WORKSHOP INFO-METRICS ACROSS THE SCIENCES (May 2, 2011) Fall 2010/Spring 2011 Institute Seminars Ximing Wu (Texas A&M) “Semiparametric Estimation of Production Functions with Shape Constraints” (with Robin Sickles) Elvira Sojli and Wing Wah Tham (Erasmus University) “The Impact of Foreign Government Investments: Sovereign Wealth Fund Investments in the U.S.” Ramo Gencay (Simon Fraser University) “Testing Jumps with Wavelets” Eric Renault (Brown) “Generalized Method of Moments with Tail Trimming” Petra Todd (U. Penn) “Structural Estimation and Policy Evaluation in Developing Countries” Robin Lumsdaine (American U.) “The Relationship between Oil Prices and Breakeven Inflation Rates” Amos Golan (American U.) “An Entropic Framework for Modelling Economies” Aman Ullah (UC, Riverside) “Estimation and Forecasting of Dynamic Conditional Covariance: A Semiparametric Multivariate Model” Hashem Pesaran (University of Cambridge) “On Identification of Bayesian DSGE Models” *Visit our website for more information. Support the Institute! Info-Metrics Spring 2011 workshop roundtable L-R: Nick Kiefer, David Wolpert, Robin Lumsdaine, and Pieter Adriaans Please visit our website for archived materials. Editors: Bethany Hardy (http://bethanyhardy.com) and Aisha Malik. Interested in providing financial support to the Institute? We are always open to new funding sources and donations. With these resources, we hope to establish more long-term fellowships for students and junior and senior researchers. We also hope to be able to expand our classes and knowledge dissemination activities. For more information, contact Aisha Malik (malik@american.edu or 202-885-3770).