New Methods - Speakers and Presentations
Saturday, October 14, 3:45-4:45 p.m.
"Multiple imputation for incomplete multilevel data with SHRIMP"
Presentation outline is available by clicking here.
Recai M. Yucel , Assistant Professor of Biostatistics, University of Massachusetts Amherst
Dr. Yucel's overall research interests are in computational statistics, applied Bayesian statistics, analysis of complex sample surveys in the presence of missing data, measurement error problems, and multivariate extensions of multilevel models. He is also interested in novel statistical thinking in applied problems such as those encountered in health services research (e.g. observational studies).
Dr. Yucel has been working on statistical techniques for analyzing incomplete data in multilevel settings.
Abstract: "Multiple imputation for incomplete multilevel data with SHRIMP"
Presentation outline is available by clicking here.
Multiple imputation is an increasingly popular method for handling missing data due to item nonresponse in surveys. When using multiple imputation, it is beneficial to reflect the data features in the imputation model. If these features involve clustering either due to sample design or study design such as longitudinal settings, one way to represent the cluster effects is via random effects in the imputation model. This idea has been developed in detail for imputing continuous variables by Schafer and Yucel (2002). Methodology for imputing categorical variables and mixtures of categorical and continuous variables is also available (Yucel, Schenker and Raghunathan, 2006), however, its use by social scientists has not been equally developed. The first part of this talk will focus on inference by multiple imputation in missing-data applications varying from continuous to mixture of continuous, categorical and "other" type of variables such as count or semi-continuous. The second part of the talk will demonstrate the use of these methods with a new program that uses Sequential Hierarchical Regression Imputation (SHRIMP).
Selected Publications:
Yucel R.M. & Schafer JL. (2002), "Computational strategies for multivariate linear-mixed models withmissing values". Journal of Computational and Graphical Statistics, Volume 11, Number 2, 437-457.
Chiu, C., Yucel, R.M., Zanutto, E. & Zaslavsky, A.M. (2005) Using matched substitutes to improve imputations for geographically linked databases. Survey Methodology, Volume 31, Number 1, 69-72.
Yucel R.M. & Zaslavsky AM. (2005) Imputation of binary responses with measurement error for treatments in health services data. Journal of the American Statistical Association, Volume 100, No. 472, 1123-1132.
Kuhlthau, K.A., Hill, K.S., Yucel, R.M. & Perrin, J.M. (2005) Financial burden for families of children with special health care needs. Maternal and Child Health Journal, Volume 9, Number 2, 207-218.
Demirtas, H., Freels, S.A., Yucel, R.M. (2007) The Plausibility of Multivariate Normality Assumption When Multiply Imputing Non-Gaussian Continuous Outcomes: A Simulation Assessment. Forthcoming in Journal of Statistical Computation and Simulation.
Yucel, R.M., Schenker, N., Raghunathan, T. (2007) Sequential hierarchical regression imputation (SHRIMP).
