Practical methods for missing data sensitivity analyses in RCTs
Learn practical methods for conducting missing data sensitivity analyses in randomised controlled trials using Stata or R.
This IN-PERSON workshop is for those statisticians analysing RCTs, with some experience with multiple imputation, and knowledge of Stata or R.
Analysing randomised controlled trials (RCTs) where some data are missing is a significant issue for statisticians which can lead to bias in estimates of clinical effectiveness. Analyses of RCTs are often carried out under the missing-at-random assumption (MAR), assuming the probability of being missing only depends on the observed data. This assumption is often questionable, and reporting sensitivity analyses under missing-not-at-random (MNAR) assumption, where the missingness may be related to the unobserved missing observation, is recommended. However, these sensitivity analyses are rarely carried out. This could be because of a lack of familiarity with the available methods or software, a reluctance by statisticians to rely on untestable assumptions, or the lack of preparation to implement such analysis, such as having access to additional data.
In this workshop, facilitated by Daisy Gaunt (UoBristol, previous NIHR Doctoral Fellow), Baptise Leurent (University College, previous NIHR Doctoral Fellow), and Suzie Cro (Imperial College London, current NIHR advanced fellowship) we aim to de-mystify and practically demonstrate available methods to conduct such sensitivity analysis. We will introduce the principles of the methods and use trial examples to demonstrate the implementation in Stata or R. We will also discuss how these methods can be planned and included in a Statistical Analysis Plan. Finally, there will be time for participants to discuss issues that occur in their trials, and for general questions and discussion.
Learn practical methods for conducting missing data sensitivity analyses in randomised controlled trials using Stata or R.
This IN-PERSON workshop is for those statisticians analysing RCTs, with some experience with multiple imputation, and knowledge of Stata or R.
Analysing randomised controlled trials (RCTs) where some data are missing is a significant issue for statisticians which can lead to bias in estimates of clinical effectiveness. Analyses of RCTs are often carried out under the missing-at-random assumption (MAR), assuming the probability of being missing only depends on the observed data. This assumption is often questionable, and reporting sensitivity analyses under missing-not-at-random (MNAR) assumption, where the missingness may be related to the unobserved missing observation, is recommended. However, these sensitivity analyses are rarely carried out. This could be because of a lack of familiarity with the available methods or software, a reluctance by statisticians to rely on untestable assumptions, or the lack of preparation to implement such analysis, such as having access to additional data.
In this workshop, facilitated by Daisy Gaunt (UoBristol, previous NIHR Doctoral Fellow), Baptise Leurent (University College, previous NIHR Doctoral Fellow), and Suzie Cro (Imperial College London, current NIHR advanced fellowship) we aim to de-mystify and practically demonstrate available methods to conduct such sensitivity analysis. We will introduce the principles of the methods and use trial examples to demonstrate the implementation in Stata or R. We will also discuss how these methods can be planned and included in a Statistical Analysis Plan. Finally, there will be time for participants to discuss issues that occur in their trials, and for general questions and discussion.