Missing Data

Even in well designed and conducted epidemiological studies, data will be missing. This may include missing observations of the exposure and under study, confounders, or the outcome.

Possible mechanisms for data being missing will be discussed, as well as their potential impact in terms of bias. Focus will be on methods to handle missing data. Examples and exercises will come from various epidemiological studies, including diagnostic, prognostic, etiologic, and therapeutic studies.

At the end of the course, you will be able to:

Missing data mechanisms

• Define and distinguish the three missing data mechanisms (MCAR, MAR, MNAR), including their key characteristics and implications for analysis.
• Determine the most plausible missing data mechanism in a given scenario or graph and justify your reasoning.
• Assess sources of missing data in a given setting and propose concrete strategies to prevent or minimize missingness.
• Distinguish the difference between sporadic and systematic missingness in the context of meta-analysis.

Imputation methods

• Explain how a given imputation method works, including how data is selected, how imputed values are calculated, and how single and multiple imputation differ.
• Assess whether the assumptions of a given imputation method are met in a specific scenario and explain the consequences if they are violated.
• Compare the effects of different imputation methods on estimators in terms of bias and efficiency.
• Identify imputation methods from graphical output.

Application

• Evaluate how missing data is handled and reported in existing research and formulate concrete recommendations for improvement.
• Design an appropriate missing data strategy for a real-world healthcare dataset, justifying the choice of imputation method based on the data context and assumptions.
• Apply R to perform missing data analysis and implement appropriate imputation strategies.
• Interpret R output from missing data analyses, drawing valid conclusions about the nature and impact of missingness.

Our courses are aimed at clinical researchers, nurses, general practitioners, and other health professionals who want to improve their skills in epidemiology, statistics and (clinical) research.

1 week, fulltime (face to face)

On the last day of the course there is a closed book exam.
Besides the exam, participants should attend at least 80% of the classes during the course.

General questions

For general questions, please contact:

Content questions

For specific questions about the course content, please contact:

Please note that this course is part of an existing program within the Graduate School of Life Sciences. Tuitition fees may alter during the year.