Session: A data driven approach for predicting non-response in longitudinal surveys: implications for missing data handling and sample representativeness

Time: Tuesday 3rd July, 15:45 - 17:15

Convenor:

Professor George Ploubidis (UCL Centre for Longitudinal Studies)

Abstract Details

Incomplete or missing data are unavoidable in longitudinal surveys. Capitalising on the richness of the Centre for Longitudinal Studies (CLS) surveys before study members attrit, we have implemented a systematic data driven approach consisting of series of multivariable regressions as well as machine learning algorithms. We will present results from the CLS Missing Data Strategy in three papers using data from the 1958 birth cohort and show that it’s possible to empirically identify predictors of non-response that maximise the plausibility of the MAR assumption and achieve estimates that under plausible assumptions are representative of their values.

The level of the session is: Accessible

Presentation details

Presentation 1

Start time: 15:45

Presentation title: A data driven approach for predicting non-response in longitudinal surveys

Presenter:

Professor George Ploubidis (UCL Centre for Longitudinal Studies)

Presentation 2

Start time: 16:10

Presentation title: Maintaining representativeness by maximising the plausibility of the MAR assumption: Evidence from the 1958 British birth cohort

Presenters:

Professor George Ploubidis (UCL Centre for Longitudinal Studies)
Mr Brian Dodgeon (UCL Centre for Longitudinal Studies)

Presentation 3

Start time: 16:35

Presentation title: The Centre for Longitudinal Studies Missing Data Strategy user guide

Presenter:

Dr Benedetta Pongiglione (UCL Centre for Longitudinal Studies)