ABSTRACT
This chapter describes some common reasons for missing values and how bias arises if missing values are ignored. It delves into missing-value mechanisms and associated statistical models. The chapter then discusses statistical models to address the impact of the different missing-value mechanisms on inference for affected outcomes and illustrates their applications under different missing data circumstances. The occurrence of missing values is a common phenomenon in modern clinical trials as well as observational studies. The chapter elaborates on some common types of missing data to help develop an appreciation of the mechanisms underlying the different types of missing values. The concept of missing values may also be applied to the counterfactual framework, a popular paradigm for studying causality, to facilitate inference, especially with observational studies. The missing at random assumption posits a mechanism that is completely determined by the observed components yi, obs of yi.
