A Practical Introduction to the Analysis of Incomplete Data — Professional Development Continuing Education Course
ASA, Biometrics Section
Incomplete data is a common complication in applied research. While most practitioners are still ignoring the missing data problem, numerous books and research articles demonstrate that dealing with it correctly is very important. Biased results and inefficient estimates are just some of the risks of incorrectly dealing with incomplete data. The purpose of this course is to demonstrate the importance of dealing correctly with incomplete data; to formulate the missing data problem and to explain the best practices to deal with this problem. We therefore, will introduce incomplete data vocabulary, ad-hoc techniques (e.g. complete case analysis, single imputation), and principled procedures (e.g. maximum likelihood, Bayesian, multiple imputation) to deal with incomplete data. We will emphasize practical implementation of the proposed strategies, including discussion of software to implement procedures for incomplete data, and the advantages and disadvantages of different missing data methodologies. At the conclusion of this course, attendees should be able to understand the complications that arise from incomplete data; able to understand and state the missing data assumptions; and to analyze incomplete data. Prerequisites: course requires knowledge of standard statistical models such as the multivariate-normal, multiple linear regression, contingency tables, as well as basic maximum likelihood for common distributions.
Instructor(s): Ofer Harel, University of Connecticut