led by Scott Oatley, PhD student at the University of Edinburgh
This training event begins with a gentle introduction into handling missing data theory, covering the necessities to engage appropriately with the practical elements of the course. By the end of the session attendees will have a solid familiarity with the different types of missingness and how each may cause potential issues surrounding bias within statistical analysis.
Following this introduction, attendees will then be taught and provided practical material in both Stata and R to duplicate specific handling missing data methods. This course splits these methods into ‘baseline’, ‘less than gold-standard’, and ‘gold-standard’ techniques. These cover: listwise deletion using OLS and maximum likelihood under a structural equation modeling (sem) framework, dummy variable adjustment, single mean/modal imputation, multiple imputation using multivariate normal regression using Markov Montecarlo methods (MCMC), multiple imputation by Chained Equations (MICE), and Full Information Maximum Likelihood (FIML).
For purposes of time, most examples and practical elements of this training use continuous outcomes. Examples are, however, provided for non-linear outcomes and their applications.
Finally, ‘gold-standard’ methods are directly compared to one another with training provided on each method including their strengths and overall weaknesses.
This training is suitable for those who are working with quantitative data and methods. Attendees should bring their own devices with either Stata or R/RStudio installed.
Accessibility information for the venue can be found here. Any other accessibility needs should be noted in your registration form.