7. Analyses under Missing-not-at-random Assumptions

  1. Michael O'Kelly1 and
  2. Bohdana Ratitch2
  1. Michael O'Kelly and
  2. Bohdana Ratitch

Published Online: 21 FEB 2014

DOI: 10.1002/9781118762516.ch7

Clinical Trials with Missing Data: A Guide for Practitioners

Clinical Trials with Missing Data: A Guide for Practitioners

How to Cite

O'Kelly, M. and Ratitch, B. (2014) Analyses under Missing-not-at-random Assumptions, in Clinical Trials with Missing Data: A Guide for Practitioners, John Wiley & Sons, Ltd, Chichester, UK. doi: 10.1002/9781118762516.ch7

Author Information

  1. 1

    Senior Strategic Biostatistics Director, Quintiles Ireland Ltd, Ireland

  2. 2

    Statistical Scientist, Quintiles, Montreal, Canada

Publication History

  1. Published Online: 21 FEB 2014
  2. Published Print: 14 MAR 2014

ISBN Information

Print ISBN: 9781118460702

Online ISBN: 9781118762516

SEARCH

Keywords:

  • missing-not-at-random (MNAR) assumptions;
  • pattern-mixture models (PMMs);
  • SAS code;
  • sensitivity analysis

Summary

This chapter discusses ways of implementing the assumption that data are not missing at random (missing-not-at-random (MNAR)), with an emphasis on using pattern-mixture models (PMMs) to do this. It first discusses the background to sensitivity analyses and PMMs. Two methods that use multiple imputation (MI) to implement sensitivity analyses with PMMs are then described and compared. The second part of the chapter is a “toolkit” indicating via SAS code fragments how to implement a variety of sensitivity analyses, using illustrative clinical trial data in three example indications. In the final section of the chapter, implementation of approaches for each of three datasets that might constitute a missing data strategy of a real study is discussed.