424 BMJ | 24 FEBRUARY 2007 | VOLUME 334 Almost all studies have some missing observations. Yet textbooks and software commonly assume that data are complete, and the topic of how to handle missing data is not often discussed outside statistics journals. There are many types of missing data and different reasons for data being missing. Both issues affect the analysis. Some examples are: (1) In a...

In this paper we discuss a possible approach to confident prediction from data containing missing values presented in a probabilistic form. To achieve this we revise and generalize the notion of credibility known in the theory of conformal prediction.

This section begins with a chapter illustrating how missing data can cloud inferences to be drawn from clinical trials – in other words, why missing data matter. Chapter 2 focuses on the mechanisms that give rise to missing data. Understanding these mechanisms is the essential background needed to understand the possible consequences of missing data. Chapter 3 discusses estimands – what is to b...