Event

Multiple imputation has become a de facto standard for handling missing data in epidemiological and clinical research. With a combination of lectures and computer practicals (Stata and R), this workshop will cover advanced topics in multiple imputation that are critical in modern research studies.

Please join us for this series of online, half-day workshops. For further information please contact vicbiostat@mcri.edu.au or series convenor Dr Cattram Nguyen cattram.nguyen@mcri.edu.au.

Introduction to multiple imputation for missing data (15 & 16 February)

An introduction to multiple imputation and the practical issues faced by researchers wishing to apply this method. In particular, the course focuses on understanding when multiple imputation is likely to produce substantial gains over a standard complete case analysis, and on the decisions faced when developing an imputation model, once it has been decided that multiple imputation is appropriate.

We provide a detailed introduction, with practical computing exercises on how to perform analyses using multiple imputation in Stata and R. The application of multiple imputation is illustrated with two case studies, in which the decisions required for implementation of the method are examined, highlighting the potential benefits as well as limitations of multiple imputation.

Sensitivity analyses to departures from the ‘missing at random’ assumption (17 February)

Standard implementations of multiple imputation are only guaranteed to provide unbiased results under the so-called “missing at random” (MAR) assumption. This roughly means that the chance of a value being missing does not depend on the value itself, given other observed data. It is therefore important to assess the plausibility of this assumption and, given that it is not testable, to perform sensitivity analyses considering scenarios where MAR does not hold (“missing not at random”—MNAR—scenarios). This workshop discusses approaches to examining the plausibility of the MAR assumption, and describes an extended multiple imputation strategy that can be used to conduct such sensitivity analyses.

Multiple Imputation for Longitudinal data (18 & 19 February)

Longitudinal studies, collecting data from individuals over time, are central in modern health and medical research. However, the prolonged observation of individuals exacerbates the risk of missing data. While multiple imputation methods for handling missing data in multiple variables are widely available in mainstream statistical software, there are important considerations, both computational and conceptual, regarding their use in the longitudinal setting. Furthermore, specialised approaches have recently been developed. Over two days, we will review the concepts and methods available for multiple imputation of longitudinal data and provide guidance on good practice.

Day 4 will provide an overview of longitudinal data analysis and methods for imputing longitudinal data in “wide” format (Stata/R). Day 5 will focus on multiple imputation methods for longitudinal data in “long” format (available in R only).

Prerequisites

Participants will require a sound working familiarity with Stata or R, and with statistics to the level of multivariable logistic regression models.