Wednesday 23rd November 2022, 13:20–13:35 (Australia/Melbourne), Zoom Breakout Room 3
When running an online probability panel complications can arise when trying to conduct concurrent cross-sectional and longitudinal surveys. These complications occur when trying to balance the needs of the two differing designs. Longitudinal designs aim to maximise the number of panellists that are surveyed repeatedly. Whereas cross-sectional designs are aiming to take a snapshot of the population, so representativeness of the sample is paramount. The presentation offers a solution that allows for both of the needs to be met at the same time, motivated by the needs of the ANUpoll, which is fielded on Life in Australia™. The ANUpoll makes extensive use of the longitudinal nature of Life in Australia™, where most other surveys fielded on Life in Australia™ are cross-sectional in nature. Our approach uses a propensity model based on diverse panel variables, including but not limited to demographic and lifestyle variables. The model allows for matching between the cross-sectional and longitudinal sample members, which are then swapped to maximise the longitudinal sample members in the longitudinal sample, without introducing bias into the cross-sectional sample.
Jack Barton is a statistician working at the Social Research Centre. Sampling and survey weighting are the main tasks Jack performs in his role as a statistician. Jack is currently undertaking a Master of Statistics and Operations Research at RMIT.