John Finamore, SRMS Publication Officer, and Michael D. Larsen, SRMS Program Chair

The ASA SRMS webinar series brings two new webinars this month. Paul Allison of the University of Pennsylvania will present “Modern Methods for Missing Data” on May 11, and Roderick Little of the U.S. Census Bureau and University of Michigan will present “The Calibrated Bayes Approach to Sample Survey Inference” May 21.

Allison’s discussion will focus on the use of maximum likelihood and multiple imputation as methods for handling missing data. Conventional methods for handling missing data are seriously flawed and often lead to biased parameter estimates, as well as serious underestimates of standard errors and p-values. Two newer methods—multiple imputation and maximum likelihood—have proven superior to more traditional missing-data methods. Maximum likelihood and multiple imputation perform well under the assumption that the data are missing at random, rather than the more severe requirement of missing completely at random. If the data are not missing at random, these two methods do well under a correctly specified model for missingness.

Allison is professor of sociology at the University of Pennsylvania, where he teaches research methods and statistics at the graduate level. He has been offering public short courses in statistics for more than 25 years.

In “The Calibrated Bayes Approach to Sample Survey Inference,” Little will describe the application of the calibrated Bayesian approach to survey sampling. Bayesian methods in statistics are increasingly popular, spurred by advances in computational power and tools. Bayesian inference provides solutions to problems that cannot be solved exactly by standard frequentist methods. Students learning the Bayesian approach will obtain new analysis tools and a deeper understanding of competing systems of statistical inference, including the frequentist approach. The objective of this webinar is to describe the application of the calibrated Bayesian approach to survey sampling, where the focus of inference is on finite population quantities. Inferences are Bayesian, but are designed to be “calibrated” in that they have good frequentist properties.

Little is associate director for research and methodology and chief scientist at the U.S. Census Bureau and Richard D. Remington Collegiate Professor of Biostatistics at the University of Michigan, where he also holds appointments in the department of statistics and Institute for Social Research.

Participants register for each webinar for a modest fee. Each registration is allowed one web and one audio connection. Multiple persons can view each registered connection. The webinars are a great way for a department, agency, classroom, or company to deliver the educational material to many people, say in a conference room equipped with a computer, projector, screen, and speakerphone. Handouts are made available a couple of business days before the presentation. During the webinar, the presenter controls the flow of the PowerPoint or similar presentation, which the audience sees on their computer screens in real time. Questions for the presenter can be submitted using the chat feature found on the webinar web page.

For more information or to register for either course, visit the SRMS webinar series web page.

Invited Sessions at JSM 2012

The Survey Research Methods Section will have the following five invited sessions at JSM 2012. Four are paper sessions, and one is an invited panel.

• International Census Coverage, Vincent T. Mule, U.S. Census Bureau
• Poverty Mapping with Complex Survey Data, Sam Hawala, U.S. Census Bureau
• Multiple Faces of Multiple Imputation: 25+ Years of Development, Innovation, and Application, Joerg Drechsler, German Institute for Employment Research
• Dual-Frame RDD vs. Address-Based Sampling: Conducting Scientific Surveys in the 21st Century, Meena Khare, National Center for Health Statistics
• Panel: Multi-Level Data Methods to Detect and Adjust for Nonresponse Bias in Sample Surveys, Tom W. Smith, NORC at the University of Chicago

One additional invited session is jointly organized with the Statistical Society of Canada: Sampling Theory and Practice: Celebrating J.N.K. Rao’s 75th Birthday

SRMS also is cosponsoring the following invited sessions of interest to many SRMS members. Check the online program in case additional sessions are added.

• Protecting Confidentiality of Complex Data: A Showcase of Statistical Activities at Federal Statistical Agencies, Benchmarking and Seasonal Adjustment
• How to Share Research Data: Views from Practitioners
• Statistical Issues in Children’s Studies
• Statistics and the 2012 Elections
• Shrinkage Estimation: Unifying Different Perspectives
• Repression of Statistics and Statisticians by the Argentine Government: Recent Developments, the Human Rights Context, and International Responses
• Are Fine Particulates Killing Californians?
• Small Population – Big Impact: Improving the Measurement of the Group Quarters Population in the American Community Survey
• Uses of Administrative Record Benchmarking in Modern Census-Taking