Seminar on Attrition and Non-Response in Longitudinal Studies of Ageing (ELSA, SHARE, HRS)
23rd March 2009
Venue: The Royal Society
Mercer Suite, 6-9 Carlton House Terrace
London, SW1Y 5AG
9.30 - 9.45 Introduction by Carli Lessof (NatCen)
Session 1. Comparison of survey design and response rates
9.45 - 10.45
Hayley Cheshire, Shaun Scholes, Mary Beth Ofstedal, and Mathis Schroeder, Response rate comparisons for HRS, ELSA, and SHARE*
This paper will highlight some of the complexities around comparing HRS/SHARE/ELSA directly. The main focus will be on the effect of differential sample and study design on response rate calculations.
*Figures for SHARE were provisional and have been excluded from presentation on web.
BREAK - 10.45 - 11.00am
Session 2. Attrition, and factors affecting non-response
This session will examine attrition and non-response, and the extent of bias created due to differences between respondents and non-respondents.
11.00 - 11.45
Anthea Tinker and Gill Mein. Factors affecting the attrition of older people in longitudinal studies
This paper will focus on the factors which affect whether older people participate in longitudinal studies beyond the first wave.
11.45 - 12.30
G. De Luca, F. Peracchi and F. Mazzonna, Determinants of the attrition process in the first two waves of SHARE
This paper will explore the determinants of the SHARE attrition process by focusing on three potential sources; mortality, survey design, and individual decisions associated with leaving a panel.
12.30 - 1.00
Hayley Cheshire, Shaun Scholes, David Hussey and Kate Cox, Factors that predict non-response in ELSA
This paper will explore the respondent characteristics associated with different patterns of participation across three waves of ELSA. The potential implications for survey practice to reduce non-response will be discussed.
LUNCH - 1.00 - 1.45pm
Session 3: Statistical techniques for dealing with non-response bias
1.45 - 2.30
Arie Kapteyn, Effects of attrition and non-response in the health and retirement study
Kapteyn et al have studied the selectivity of attrition in the original 1992 HRS cohort and its impact on the representativeness of the sample. This paper proposes several tests of the extent of attrition bias, and studies the impact of propensity weighting-a statistical approach to correcting for selective attrition-on analytic results.
2.30 - 3.15
James Banks, Alastair Muriel and James P Smith, "Adjusting for differential attrition when estimating dynamic processes for health and socioeconomic status in HRS and ELSA."
This paper will compare the patterns of longitudinal non-response across the health and SES distribution in the HRS and ELSA samples. The analysis will consider the extent of differences in baseline health and socioeconomic characteristics between those who remained in the study and those who subsequently left the study.
BREAK - 3.15 - 3.30pm
3.30 - 4.15
David Weir, Jessica Faul and Kenneth Langa, Jessica Faul, Approaches to assessing and correcting for bias in distributions of cognitive ability due to biased non-response in HRS and ELSA
This paper compares rates of longitudinal non-response in HRS and ELSA, and the magnitude of cognition bias that results, using data internal to both surveys. We show the value of proxy interviewing for offsetting bias. For the HRS, we then conduct further work using external data from Medicare records to assess bias and propose corrections.
4.15 - 4.45
Harvey Goldstein, Handling attrition and non-response in longitudinal data
This paper will discuss ways in which some investigators have dealt with non-response through modelling, weighting and imputation. A multiple imputation (MI) strategy will be discussed that can be applied to complex multilevel data.
Session 4. Discussion about the day
4.45 - 5.30
This discussion will provide an opportunity to summarize lessons learned, and to establish priorities for further research and for modifications to survey practice with respect to longitudinal participation.
5.30 Sum-up and finish
|
