Multiple Imputation - A Practical Introduction
Incomplete dataset arise often: for example in clinical trials because of patient non-compliance and in observational studies because of unobserved explanatory variables. Imputation is the process of replacing missing data with a probable value based on other available information, so as to preserve all cases.
This course explores both imputation process and the analysis of a dataset that has undergone imputation. This course emphasises the practical aspects of multiple imputation but the underlying theory is also outlined. The statistical package Stata is used to illustrate the methodologies in the presentations and for practical work. R and SAS may also be used for practical work.
Who Should Attend?
Scientists and analysts engaged in statistical analysis of data, who have no experience of multiple imputation. A working knowledge of statistical modelling (e.g. logistic regression) is required, as well as of one of the statistics packages listed above.
How You Will Benefit
This course will give a thorough introduction to Multiple Imputation by chained equations for multiple incomplete variables, both numerical and categorical. Various imputation models will be explored, as well as the presentation of results from a Multiple Imputation analysis.
What Do We Cover?
- Impact of missing data; missing data mechanisms.
- Multiple imputation for an incomplete single variable
- Chained equations for imputing multiple incomplete variables
- Model building using multiply imputed data; combining estimates
- A checklist for reporting results.
This course has practical exercises written for: R, SAS, Stata