JSM Short Courses, Other Meetings Opportunities Announced

Edited by Songthip Ounpraseuth, Biometrics Section Publications Officer

The Biometrics Section is proud to sponsor the following two short courses during JSM 2011 in Miami Beach, Florida:

• Generalized Linear Mixed Models

  Instructor: Charles McCulloch, University of California, San Francisco

  The class of generalized linear mixed models (GLMMs) is a broad class of statistical models generalizing both linear mixed models (LMMs) and generalized linear models (GLMs). As such, it is capable of accommodating nonlinear responses, correlated data, and non-normal distributions. This makes it useful in practice. For example, GLMMs give a natural way to specify a correlated data model for binary data.

  This course will briefly review the concepts of linear mixed models and the use of random effects as well as the modeling strategy behind generalized linear models. From these two classes of models, generalized linear mixed models will be developed. A series of examples will be considered to develop intuition about how to specify these models in real situations. Next, features of generalized linear mixed models will be developed and strategies for fitting the models to data will be described and contrasted with approaches such as generalized estimating equations. A series of case studies will be used to illustrate the practical use of these models. The focus in the course will be on approaches to modeling, methods of estimation and inference, and available software.

• Semiparametric Theory and Missing Data

  Instructor: Anastasios A. Tsiatis, North Carolina State University

  This course will be a full-day short course broken down into two distinct sessions. The morning session will introduce the theory and methods for semiparametric models assuming there are no missing data (i.e., the full-data problem). Using ideas developed in the morning session, the afternoon session will discuss how to extend these ideas to missing data problems and show how this leads to an augmented inverse probability weighted complete-case AIPWCC estimators.

  In the morning session, semiparametric models will be formally defined and some of the theoretical developments for estimators of the parameters will be reviewed. Most practical estimators for either parametric or semiparametric models are asymptotically linear in the sense that they can be approximated by a sum of iid random variables referred to as the influence function. We show that the large sample properties of such an estimator are directly related to its influence function; specifically, the asymptotic variance of the estimator is equal to the variance of its influence function.

  The beauty of semiparametric theory is that influence functions can be viewed as geometric objects (i.e., vectors in a linear space where distance away from the origin is related to the variance of the influence function). As such, estimators whose influence functions have small variance, that is, short distance, are desirable. This geometric perspective allows us to visualize and construct better estimators using projections onto appropriate linear subspaces and characterize the most efficient estimator, that is, the estimator with the smallest asymptotic variance. We first illustrate how this geometric theory applies to finite-dimensional parametric models and then show how these ideas can be generalized to semiparametric models. Several pedagogical examples will be used to illustrate the theoretical developments.

  In the afternoon session, these ideas will be applied to missing-data and, more generally, to coarsened-data semiparametric problems. We will buy ativan china begin by introducing different missing data mechanisms, such as missing completely at random (MCAR), missing at random (MAR), and nonmissing at random (NMAR). We will focus primarily on problems where missing are MAR. The geometric ideas for semiparametric full-data models will be extended to missing-data models and will lead to a deeper understanding of the underlying theory for missing data. Methods for estimating parameters in as efficient a manner as possible while still being feasible will be emphasized. As we will show, this theory leads naturally to AIPWCC estimators. Examples will be given to illustrate the methods.

  The course requires that participants have taken an advanced course in inference and probability and have a good understanding of large sample theory (i.e., a good understanding of convergence in probability and convergence in distribution). It also would be helpful if participants were exposed to functional analysis and Hilbert spaces, although this is not required.

ENAR 2012

It is time to think about invited sessions for ENAR 2012, which will be held April 1–4 in Washington, DC. Anyone who is interested in organizing an invited session or who has ideas for one should contact the section’s 2011 program chair, Jason Fine, at jfine@bios.unc.edu.

A typical session consists of three 30-minute talks followed by a 30‑minute discussion or four 30-minute talks. June 11 is the deadline for proposals. It is best if you have a well-defined topic and have commitments from participants by June 11. The more detailed the proposal, the better the chances it will be selected in this highly competitive process.

JSM 2012

It’s also time to start thinking about invited sessions for next year’s Joint Statistical Meetings, which will be held July 28 to August 2 in San Diego, California. Anyone interested in organizing an invited session or who has an idea for one should contact the section’s 2011 program chair, Tianxi Cai, at tcai@hsph.harvard.edu.

A typical invited session consists of three 30-minute talks followed by a 10-minute invited discussion and 10 minutes of floor discussion. However, other formats are possible. The 2011 program is a good source for examples. Remember, the most mature ideas will have an advantage in competing for the limited number of slots, so it’s best to have your ideas in final form by the middle of June. The Biometrics Section will have at least four invited sessions, but if we generate enough good ideas, we will be able to compete for additional slots.

Ideas for short courses should be sent to the section’s 2011–2012 continuing education chair, Annie Qu, at anniequ@illinois.edu.

Special Announcement

CHANCE magazine (an ASA publication) features a regular column called “Here’s to your Health.” The column publishes interesting statistical issues in health-related topics presented at a basic level. The magazine is always looking for researchers to write short papers for the column and would like to invite you to write a paper. To provide a flavor for the types of articles published, here are some recent column topics:

• Statistical issues in pharmacogenetics
• A re-examination of the 1918 flu pandemic
• Application of PET scanning to diagnosing mental illness
• Assessing cardiovascular risk for women on estrogen replacement therapy

If you are interested in writing for the column or have questions, email Mark Glickman at mg@bu.edu.