Assessing and adjusting for missing data in longitudinal studies
20th April 2016
CMIST, Humanities Bridgeford Street
It is intended that students attending the course will be able to assess the implications of non-response for different kinds of datasets that they might use. Students new to longitudinal data should consider taking the CMIST short course running immediately before this one. The course will not cover multiple imputation in any detail and students particularly interested in this topic should take one of the other specialist NCRM courses.
The course begins with a general overview of methods for handling missing data. It is followed by an introduction to response propensity models and their use in constructing non-response weights; this is linked to a practical session using data from the National Child Development Study (NCDS). After lunch, there is a session on how to estimate R (for representativity) indicators and Receiver Operating Characteristic (ROC) curves for assessing bias and prediction; this is linked to a practical session using data from the Millennium Cohort Study (MCS). The day concludes with a session on assessing missingness with repeated measures data.
For more information and to book