Houcan Zhang 张厚粲, Beijing Normal University, Beijing, China
Houcan Zhang is a Professor of Psychology and a Member of the University Administration Board of Beijing Normal University. She was educated at Fu-Ren University, Beijing China, 1948 and was conferred an Honorary D.Sc of Fu-Jen University, Taiwan China, 2005. Her research interests are experimental psychology, human cognition, and educational and psychological measurement. She took charge of the standardization of the second and fourth editions of the Wechsler Intelligence Scale for Children (WISC) in China, and developed the Intelligence Scale for Chinese Children 3-6. Professor Zhang held the offices of Vice-President and Member of the Executive Committee of the International Union of Psychological Science (1996－2004), Council Member of the International Test Commission (1990－1994), President of the National Association of Educational Measurement and Statistics, Vice-President of the Chinese Psychological Society, Director of the Educational Psychology Division of the National Steering Committee of Educational Science, and Dean of the Department of Psychology at Beijing Normal University. Aside from her academic roles, Houcan Zhang has served as Counselor of the State Council of the People’s Republic of China (1988-2005) and Member of the Chinese People’s Political Consultative Conference (1993-2002). Houcan Zhang was named Distinguished Personality of the City of Beijing. She was awarded the Worker of Excellence in Science and Technology by the National Association of Science and Technology in 2005, the Life-long Achievement Award by the Chinese Psychological Society in 2001, and the Life-long Achievement Award on Educational Research in 2011.
Psychometrics in China: History and its development
The idea of individual difference and its measurement has a long tradition in China. About 3000 years ago, Confucius noticed individual differences among people and Mencius proposed that these differences can be measured. In the 7th century, China established the Civil Service Examination for personnel selection, which continued for more than 1000 years till early 20th century. The method of testing had also been spread to many other countries like Europe. China is thus called the birthplace of test. Psychometrics, as a branch of psychological science, was introduced into China at the early 20th century. It was mainly disseminated in the field of education and went through a rather prosperous stage. Unfortunately, psychology was criticized and testing was forbidden because its belief in individual differences was in conflict with contemporary ideology in the 1950s. And it was totally denied as a discipline of science during the Cultural Revolution (1965-1976). After the Cultural Revolution, psychology was revived. As for psychometrics, we started the first nationwide workshop on testing and psychological statistics in 1980 to train teachers of most universities. Then psychometrics regained its reputation through its successful application in the National College Entrance Examination reform, which promoted the development of psychometrics in education. Further, greater demand for mental tests emerged in the field of clinics and counseling. As with the development of society, personnel selection reform also took place. Psychometrics was viewed as an invaluable tool for informing decision-making about employee and related organizations. It also contributed a lot to the field of counseling. In the late 1990s, psychology, especially psychometrics, became well-recognized and welcomed by the public. At the same time, we psychologists were aware of cultural difference and the urgent need for development of native tests. In order to monitor the basic education quality and students' mental health to further inform education policy making, the Ministry of Education developed a comprehensive scale for educational assessment. Besides, the National Education Examinations Authority cooperated with the OECD and took a part in the PISA Program since 2006, and found many of its methods can be applied in China. The Shanghai Institute of Education led the same PISA program and gained excellent results. To follow recent advanced development in psychometrics, besides its application, many studies in the field of Cognitive Diagnosis Theory, Computerized Adaptive Testing and advanced statistics can be found as well.
Yutaka Kano is a professor of statistics and psychometrics at Osaka University. He received his Ph.D. in mathematical statistics from Osaka University in 1986. Since then, he has held academic positions at departments of mathematics, mathematical sciences and human science at several universities. He has served as associate editor of Psychometrika and the Journal of Multivariate Analysis, among others. He has contributed particularly to areas of asymptotic theory, structural equation modeling, reliability theory, statistical causal inference and analysis of incomplete data. The first International Meeting of the Psychometric Society in Asia was held in Osaka, Japan in 2001, where he was chair of the local organizing committee.
Developments in multivariate missing data analysis
It is well known that missing data can cause a serious bias in statistical inference if they are ignored. In this talk, we will provide two recent developments on missing data analysis. One is a proposal of an analysis method of data with a huge amount of missing values, say 90 percent of data points. This situation can happen if subjects can choose items to be responded in a questionnaire study. We apply it to make factor analysis of a real data set in an Internet survey research. The second is a theoretical study on effects of inclusion of auxiliary variables to reduce a bias of the MLE without any missing-data mechanism for NMAR missingness. Recently the method for missing data has been reported to be very useful. Some Mote Carlo studies in the literature have shown that inclusion of auxiliary variables can successfully reduce the bias greatly. We first provide a general framework to study the bias of the MLE for NMAR missingness, and then apply it to evaluate effects of auxiliary variables. Surprisingly there are cases where inclusion of auxiliary variables enlarges the bias.
Simon Wood, University of Bath, Bath, UK
Simon Wood is a professor of Statistics at the University of Bath, with particular research interests in statistical computing, all aspects of the theory, computation and application of additive smooth models and environmental modelling. He is author of the R recommended package 'mgcv', for generalized additive mixed modelling, as well as a number of research papers on GAMs and two books: "Generalized Additive Models: an introduction with R" and the forthcoming "Core Statistics". Originally trained in physics, he then moved into theoretical ecology (University of Strathclyde and Imperial College, London), which led naturally on to statistics (Universities of St Andrews, Glasgow and Bath). He currently holds a 5 year ‘established career fellowship’ from the UK Engineering and Physical Sciences Research Council, enabling him to concentrate on research in the area of smooth modelling.
Modelling with smooth functions
Deciding which variables are potential predictors in a regression model is often much easier than deciding how they should enter the model. Generalized additive models and their extensions address this imbalance by allowing models to be specified in terms of smooth functions of predictors, where the functions are then the object of inference. This talk will review the rich variety of smooth model components that can be used to construct such models, and how they can reliably be estimated, including estimation of the degree of smoothing. By taking an empirical Bayesian approach based on reduced rank splines and using Laplace approximate marginal likelihood to estimate smoothing parameters, a quite widely applicable framework can be constructed encompassing random effects, functional predictors and response distributions well beyond the usual single parameter exponential family, including multivariate and location-scale smooth regression models. The talk will cover practical application of this framework and its software implementation in R.
Lawrence Hubert, University of Illinois at Urbana Champaign, IL, USA
2015 Psychometric Society Career Award for Lifetime Achievement
Lawrence Hubert is the Lyle H. Lanier Professor of Psychology and Professor of Statistics and Educational Psychology at the University of Illinois at Urbana-Champaign. He received his PhD in Mathematical Studies in the Educational Processes from Stanford in 1971 and has held positions as Professor of Educational Psychology at the University of Wisconsin, Madison, and the University of California, Santa Barbara. Hubert’s research program has concentrated on the development of exploratory methods for data representation in the behavioral sciences, emphasizing cluster analysis, a range of spatially oriented multidimensional scaling techniques, and several network representation procedures. Much of this work on Combinatorial Data Analysis is summarized in two research monographs with the Society of Industrial and Applied Mathematics (with co-authors P. Arabie and J. Meulman): Combinatorial data analysis: Optimization by dynamic programming (2001); The structural representation of proximity matrices with MATLAB (2006). More recently, Hubert delved into the interface between statistics and ethical practice. A 2013 CRC Press book with Howard Wainer, A statistical guide for the ethically perplexed, documents this work. Hubert has been President of the Psychometric Society and of the Classification Society of North America and an elected Foreign Member of the Royal Netherlands Academy of Arts and Sciences. His editorships have included Psychometrika and Journal of Educational Statistics. Hubert is Fellow of the American Statistical Association, American Psychological Association, Association for Psychological Science, the American Association for the Advancement of Science, and the American Educational Research Association. In 2009, he received the Jacob Cohen Award for Distinguished Contributions to Teaching and Mentoring from Division 5 of the American Psychological Association.
An old tale of two computational approaches to regression:
Updated for the 21st century
Two different approaches from the early 1900s to the computational task of multiple regression are attributed to a psychometrician, Truman Lee Kelley, and a major political figure, Henry A. Wallace. The Kelley approach would today be called an alternating least-squares strategy; a convergence criterion is set but there is no fixed number of operations. For Wallace, it was the 1925 publication, Correlation and Machine Calculation, with George Snedecor that detailed the computational steps for solving the normal equations through Gaussian elimination (or the Doolittle method); the approach was algorithmic with a fixed number of operations. The Kelley iterative alternating least-squares approach is alive and well today in the contemporary psychometric literature in the form of the Kaczmarz-Dykstra iterative projection strategy for solving linear inequality constrained least-squares tasks. It has been used to find and fit unidimensional scales (linear and circular), multidimensional city-block scalings, ultrametric and additive trees, and various other order-constrained structures all fit to given proximity matrices.
Sy-Miin Chow, Pennsylvania State University, PA, USA
2015 Psychometric Society Early Career Award
Sy-Miin Chow earned her PhD in quantitative psychology from the University of Virginia in 2004. She is now an Associate Professor in the Department of Human Development and Family Studies at the Pennsylvania State University. Dr. Chow is a recipient of the Humboldt Research Fellowship from the Alexander von Humboldt Foundation in Germany and the Cattell Early Career Award from the Society of Multivariate Experimental Psychology (SMEP). Dr. Chow’s research focuses on the development and adaptation of modeling and analysis tools that are suited to evaluating linear and nonlinear dynamical systems models, including differential/difference equation models, longitudinal structural equation models, state-space modeling techniques and Bayesian dynamic models. She has served as a member of the editorial advisory board of SMEP and the associate editor of Psychometrika and Psychological Methods.
Fitting and evaluating nonlinear dynamical systems models:
Current techniques and unresolved challenges
Dynamical systems are systems that change over time such that their current states are somehow dependent upon their previous states. Change concepts described in most dynamical systems models are by no means novel to social and behavioral scientists, but most applications of dynamic modeling techniques in these disciplines are grounded on linear theories of change. In this talk, I will illustrate the potential utility of nonlinear dynamical systems models using empirical examples drawn from the behavioral sciences. Current frequentist and Bayesian techniques for performing exploration, estimation and diagnostics of nonlinear dynamical systems models, as well as challenges and unresolved issues in utilizing these techniques, will be reviewed.
Michelle LaMar, Educational Testing Service, NJ, USA
2015 Psychometric Society Dissertation Prize
Michelle LaMar is an Associate Research Scientist in the Center for Advanced Psychometrics at Educational Testing Service. Her current research focuses on the development of psychometric models appropriate for use with complex assessment tasks such as simulations or games. She is particularly interested in modeling task-process data using dynamic cognitive models to enable valid inference about multiple layers of student cognition. Michelle received her Ph.D. in educational measurement from the University of California, Berkeley in 2014, studying under Sophia Rabe-Hesketh. Prior to her doctoral work, Michelle spent 18 years in software engineering, specializing in educational simulations, authoring tools, and natural-language parsing.
Cognitive models for understanding student thinking in complex tasks
Complex tasks require complex cognition for which straight-forward unidimensional models are unlikely to be a good approximation. As current educational reform emphasizes integrated performance tasks, and modern technology offers rich streams of performance data, the use of more complex cognitive models in psychometrics may be increasingly useful and feasible. Recent work will be presented using a cognitive model borrowed from computer science, the Markov decision process (MDP), as the basis for a flexible psychometric framework. The MDP measurement model enables inferences about student goals, beliefs, and strategic thinking, either individually or in combination. Issues of parameter estimation and interpretation are explored through simulation, while empirical studies illustrate the validity, practicality and limitations of the approach.
Sophia Rabe-Hesketh, University of California, Berkeley, CA, USA
President of the Psychometric Society, 2014-2015
Sophia Rabe-Hesketh is a Professor of Educational Statistics and Biostatistics at the University of California, Berkeley. She was previously Professor of Social Statistics at the University of London where she also completed her PhD in Physics. Her research interests include latent variable, multilevel and longitudinal modeling, estimation methods, and missing data. She has published over 100 peer-reviewed journal articles and coauthored six books. Her h-index in Google Scholar is 57, and her gllamm software has been used in over 700 different journals. She is a fellow of the American Statistical Association and has been elected to the National Academy of Education and the International Statistical Institute.