Phil Kott

Happy very belated New Year! Since this message is going out a good bit later than usual for reasons beyond my control, I needed to make a small change to the annual salutation before I jumped into the customary outline of some of what the Survey Research Methods Section (SRMS) will be doing over the next year.

Much of SRMS’ activity in 2014 will be centered on the Joint Statistical Meetings (JSM), to be held in Boston, Massachusetts, in August. SRMS will again sponsor many activities, including short courses and a variety of technical sessions.

In 2013, our section was one of the five to experiment with JSM SPEED sessions, in which each contributor gave a five-minute oral presentation followed later that day by a poster session. SPEED sessions will appear again in 2014 and may be the future of JSM.

If you will be in Boston for JSM, check the online program for the SRMS business meeting and mixer, normally held Wednesday evening. Section members and nonmembers are welcome to attend to learn more about the section’s activities and, if so inclined, provide ideas and input into future activities. Best of all may be the free food and drink available while socializing with colleagues. Nonmembers can easily duck out before the formal meeting starts. Members too, but many will find the meeting itself of surprising interest.

Besides the short courses and technical sessions at JSM, SRMS is again planning for a series of webinars throughout the year. The webinars are announced to the SRMS membership via email and in the section newsletter, and information about upcoming webinars also is provided on the section website.

A valuable source of information is the section’s electronic mailing list hosted by the University of Maryland. I have used it on occasion for lazy man’s research. It you don’t know about the mailing list, visit the SRMS website.

The ASA co-sponsored Journal of Survey Statistics and Methodology (JSSAM) began publishing last year. It is available for free to SRMS members. JSSAM aims “to publish cutting-edge scholarly articles on statistical and methodological issues for sample surveys, censuses, administrative record systems, and other related data.” There is also a section for e-letters on the journal’s website.

One of the section’s most important activities is fostering, supporting, and encouraging scholarly work by students entering the field of survey research. Together with the Social Statistics Section and Government Statistics Section, SRMS co-sponsors a student travel award competition that provides the awardees with funds to support their travel to JSM.

Here’s hoping that our section and we have another productive year!

Short Course on Calibration Weighting in Survey Sampling

The ASA’s Survey Research Methods Section is sponsoring a half-day continuing education short course at this year’s JSM titled “Calibration Weighting in Survey Sampling.” The course will be taught by Phil Kott of RTI International. Information about how to register for the course is available on the JSM website.

Calibration weighting involves a mild adjustment of probability sampling weights that forces the weighted totals for a set of calibration (benchmark) variables to equal values determined using more complete information from the frame, target population, or a larger sample. Its use can increase the efficiency of survey estimates and adjust for frame coverage errors and unit nonresponse. This short course is composed of three modules of increasing complexity:

• The first module provides a broad overview of the topic. It assumes knowledge of survey sampling at the level of Lohr’s Sampling Design and Analysis.
• The second module discusses the roles of the linear prediction models and probability-sampling theory in calibration weighting. A treatment of optimal and pseudo-optimal calibration is followed by a discussion of double protection from potential biases due to unit nonresponse or coverage errors. The potential use of calibration weighting when nonrespondents are not missing at random is introduced. Some familiarity with derivatives and linear algebra is needed.
• The third module discusses large sample variance estimation for which previous exposure to asymptotics would be helpful, although concepts are stressed over rigorous proofs.

A motivating example grounds the theory throughout by displaying the numerical impact of alternative approaches. Handouts of slides and SUDAAN 11 code used to produce numerical examples will be provided to students. Note that this is not a course in SUDAAN procedures, so there will be no explicit discussion of the code in class.

The learning objectives for the course include understanding what calibration weighting is and how to use it, how calibration weighting relates to other methods of weight adjustment, how it can be used to increase the efficiency of estimated means and totals and to reduce selection biases due to nonresponse or coverage errors assuming either a selection or outcome model, how it can be used when nonresponse is not missing at random, how to measure the standard errors of resulting survey estimates properly, what some of the limitations of calibration weighting are, and what is yet unknown.