From CRMDA

Modern Missing Data Treatments and Designs
Five-day Course • June 11 –15, 2012• Lawrence, Kansas

Presented by the Quantitative Training Program of the Center for Research Methods and Analysis at the University of Kansas

Institute Overview

Topics include the mechanisms of missing data (MCAR, MAR, MNAR), modern unbiased estimation of missingness (i.e., maximum likelihood and multiple imputation), planned missing data designs, and special procedures for large, longitudinal, and nested data.

The course is designed for the applied researcher to gain full command of the arsenal of modern missing data techniques that enable the researcher to recover power and accuracy from data with missingness. The first part of the course will involve in-depth coverage of types of missing data (e.g., What does it mean for data to be "missing at random"?), how and why missingness arises, and what can be done to prevent unplanned missingness. Researchers will learn about the benefits of planned missing designs (i.e., designs where a certain percentage of data are randomly selected to be missing) in both cross-sectional and longitudinal studies and how to implement these in their research. The third part of the course will focus on modern methods of dealing with missing at random data, including maximum likelihood and multiple imputation approaches, as well as models for dealing with missing not at random. The course will cover the implementation of these techniques in several pieces of popular software, including SAS, Mplus, R, and SPSS. The benefits and drawbacks of each technique will be presented. Finally, the last part of the course will cover advanced statistical issues to consider when missing data are present, for example, robust corrections for model estimates when nonnormality and missing data co-occur. Ample time will be available for researchers to practice these methods and apply them to their own research under the guidance of experts in the field.

Objectives

The institute on Modern Missing Data Analysis will enable participants to:

• Distinguish between different kinds of missing data (e.g., missing at random) and understand the problems posed by missing data to data analysis.
• Design research to minimize the amount of unplanned and nonrandom missing data.
• Design research that utilizes planned missing data to maximize power given a fixed number of data.
• Understand the limitations of traditional approaches to missing data (e.g., listwise deletion and regression imputation).
• Implement modern missing data methods (e.g., FIML, EM, and multiple imputation) with a variety of software (Mplus, SAS, SPSS, and R).
• Argue convincingly (e.g., to reviewers) that modern missing data methods do not constitute cheating!

Audience

This course is intended for researchers in a variety of fields who want to gain a deeper conceptual, mathematical, and practical understanding of modern missing data analysis tools. The course assumes some basic understanding of research methods and statistics, but will be largely self-contained. The course may use examples from regression and basic structural equation models, for which an understanding of regression and a familiarity with SEM (e.g., confirmatory factor analysis models and growth curve models) would be useful. However, different data analysis model is not the primary focus of this course.

Instructors

Wei Wu, Ph.D., is assistant professor of psychology at the University of Kansas. Wei received her PhD in Quantitative Psychology in 2008 from Arizona State University. Her research focuses on longitudinal data analysis and missing data methods.

Mijke Rhemtulla, Ph.D., is a postdoctoral researcher at the Center for Research Methods and Data Analysis at the University of Kansas. Mijke received her PhD from the University of British Columbia in 2010 in Developmental Psychology.

Software and Computer Support

Participants should bring a laptop computer. The following software programs will be covered. Guest access to “EduStat” server in the Center for Research Methods and Data Analysis at the University of Kansas will be provided for the use of the programs.

1. SAS.
2. SPSS. A 14-day trial version of SPSS may be found at http://www14.software.ibm.com/download/data/web/en_US/trialprograms/W110742E06714B29.html
3. Mplus. A demo version can be downloaded free from http://www.statmodel.com/demo.shtml. This site also has information about pricing.
4. R. R is free, and may be downloaded here: http://www.r-project.org/

Literature

Enders, C. K. (2010). Applied missing data analysis. New York: The Guilford Press.
Schafer, J., & Graham, J. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7, 147-177.
Graham, J. (2009). Missing data analysis: Making it work in the real world. Annual Review of Psychology, 60, 549-576.
Rubin, D. (1996). Multiple imputation after 18+ years. Journal of the American Statistical Association, 91, 473-489.
Graham, J. W., Taylor, B. J., Olchowski, A. E., & Cumsille, P. E. (2006). Planned missing data designs in psychological research. Psychological methods, 11, 323-343.
Little, R. J. A., & Rubin, D. B. (2002). Statistical analysis with missing data. Wiley-Interscience, Hoboken, NJ.

Syllabus

Modern Missing Data Treatments and Designs
Wei Wu and Mijke Rhemtulla
University of Kansas

* Our group lunch is included in the Holiday Inn room rate and tickets for the group lunch are also available for purchase for those not staying at the Holiday Inn. There are other lunch options in Lawrence as well.

Contact Information

For information on course content, contact Wei Wu or Mijke Rhemtulla
A full list of prices and fellowship opportunities for this course and all the courses offered at this year's Summer Institutes in Statistics can be found on the Fees and Registration Page.