For the social sciences, collecting complete data can be a significant challenge. Item non-response can occur for a multitude of reasons, and particularly with politically or personally sensitive questions. Non-response to survey questions can result in the loss of statistical power, and even more worryingly can distort inferences about the population under study and lead to invalid conclusions.
In this one-day workshop we will look at the underlying mechanisms of missing data and how these can impact on statistical inferences. Next, we will discuss the ad-hoc methods that have been used in the past to address missing data and why these should be seen as wholly inappropriate for published work. Lastly, we will introduce the broad approach of multiple imputation - specifically focusing on joint modelling and the fully conditional specification methods.
The workshop assumes no prior knowledge of missing data analysis techniques, although participants should be familiar with multiple regression. Throughout the practical sessions the software of instruction will be R. R has a number of well-designed packages to carry out missing data analysis – Amelia, mi, mice, Jomo etc. However, instructional materials will also be provided for other commonly used software (e.g. STATA) where possible.
Course convenor: Dr Michael Adkins
PhD students, Masters in Social Science Research Methods students, and Social Sciences colleagues are all welcome to attend.