Description
Missing data are a ubiquitous feature of nearly all research applications, arising through participant non-response, attrition, and sometimes even by design. Failure to account appropriately for missing values when conducting statistical analyses can result in badly biased estimates and incorrect inferences about the relationships under study. Modern Missing Data Analysis focuses on three optimal approaches for addressing missing data that can be applied across a variety of research settings: maximum likelihood, Bayesian estimation, and multiple imputation. Participants will learn details about the classic versions of these methods as well as state-of-the-art extensions that have been developed in the last five years. These procedures are advantageous because they use all available data and make realistic assumptions about the cause of missingness; estimates and significance tests are therefore valid in a broader range of situations than historical methods such as deleting incomplete data records. The purpose of this course is to provide participants with foundational knowledge about maximum likelihood, Bayesian estimation, and multiple imputation. Equally important, participants will learn how to select an appropriate missing data handling method and apply it to complex, real-world data sets. To this end, presentations will include a mix of theoretical information, practical tips, and computer applications.
