Our webinar series, sponsored jointly by the American Statistical Association’s Committee on International Relations in Statistics (CIRS) and by Statistics Without Borders (SWB), provides introductory lectures by experts on important topics of current interest and is aimed at an international audience.

Missing data are a common problem in statistics; examples include unit and item nonresponse in surveys, attrition in longitudinal data sets, and missing data arising from noncompliance to treatments in clinical trials. We will review some approaches to handling the problem. Topics include (a) pros and cons of common methods, specifically the analysis of the complete cases, nonresponse weighting and extensions, maximum likelihood, Bayes, and multiple imputation; (b) approaches to missing data when the data are potentially missing not at random; (c) subsample ignorable likelihood approaches for regression with missing data, which address particular missing not at random mechanisms by selectively omitting data; and (d) causal inference under noncompliance to treatments as a missing data problem.