Integrating survey data collected in different ways: how can we produce robust population estimates?
Speaker(s):
Bio: Andrew is Professor of Trials in Global Health jointly in the Institute for Global Health, and the MRC Clinical Trials Unit. He has been employed at UCL since 1994, during which time he completed a part-time PhD in statistics. Andrew's core expertise is in medical statistics. He is the lead statistician for several ongoing studies in the UK and overseas, including the National Survey of Sexual Attitudes and Lifestyles (Natsal). He has a research interest in cluster randomised trials, survey methods and missing data and has previously published on the challenges of trying to weight web panel data to be used alongside Natsal.
Abstract:
Face-to-face probability sample survey fieldwork is increasingly challenging and expensive, and many surveys are looking into alternative data collection options, including using multiple methods for data collection. Examples include using both probability and non-probability sampling frames, and/or modes of data collection (face-to-face and remote modes, e.g., online and telephone interviews). But how do survey researchers combine the resulting data to produce robust population estimates of prevalence or association? Should, and if so, how can the relative quality and inherent biases of the different data sources be taken into account? This panel discussion will bring together experts from across major survey organisations and academia to discuss the challenges of integrating data and offer their thoughts as to how surveys might address this problem . The panel discussion will be hosted by researchers from Natsal (www.natsal.ac.uk), one study facing this challenge.