Theoretical Sampling Design Options for a New Birth Cohort
Lead Research Organisation:
University of Manchester
Department Name: Social Sciences
Abstract
The 2017 Longitudinal Studies Strategic Review (Davis-Kean, Chambers, Davidson, Kleinert, Ren and Tang, 2018) made a series of recommendations on innovative ways to enhance and to invest in opportunities for longitudinal research in the UK. One of the key recommendations to the ESRC was to commission a new birth cohort with an accelerated longitudinal design. The accelerated longitudinal design engages multiple cohorts simultaneously, each one starting at a different age and then following a 'suite' of aligned cohort groupings over time. In this proposal we investigate high-level design options and feasibility for an accelerated longitudinal design in the UK for a cohort of individuals starting soon after birth and a second or multiple cohorts starting in childhood and following each one over time, at least to age 18. The 2017 Longitudinal Studies Strategic Review mentions 'pivotal age periods (to be determined)'. Given the aim of sampling from an 'administrative spine' which in practice does not exist, we can assume that one age point would be 0, e.g. from the registry of births. Another key administrative source for children is the School Census and it is natural to take the age of 5 (the start of compulsory schooling) as an additional age point. A third age point could be the age of 16 (the end of compulsory schooling). We note that the issue of following a cohort during pregnancy may be problematic due to a lack of a sampling frame.
More specifically, we investigate theoretical aspects of an accelerated longitudinal design versus a traditional single cohort with respect to: statistical power and precision of estimates to be able to examine key research questions; consideration of geographical aspects and adequate sample sizes at the UK national and regional level and in vulnerable subgroups (including booster sampling); statistical analysis under multiple cohorts; comparability of measures across cohorts and across sweeps; missing data and attrition.
More specifically, we investigate theoretical aspects of an accelerated longitudinal design versus a traditional single cohort with respect to: statistical power and precision of estimates to be able to examine key research questions; consideration of geographical aspects and adequate sample sizes at the UK national and regional level and in vulnerable subgroups (including booster sampling); statistical analysis under multiple cohorts; comparability of measures across cohorts and across sweeps; missing data and attrition.
Planned Impact
The results of this research will be disseminated through presentations at relevant institutions, such as the Centre of Longitudinal Studies at UCL Institute of Education. In addition, an academic journal paper will be prepared for publication in JRSS A.
Organisations
Publications
| Description | The research project: Theoretical Sampling Design Options for a New Birth Cohort: an Accelerated Longitudinal Design Perspective, was invited research from the ESRC following the 2017 Longitudinal Studies Strategic Review by Davis-Kean, Chambers, Davidson, Kleinert, Ren and Tang (2018). One of the recommendations in the strategic review was to examine the potential for an accelerated longitudinal design to reduce the length of time needed for a birth cohort study. The deliverable of this research was a report. The report included a comprehensive and extensive literature review of accelerated longitudinal designs with descriptions of international studies that focus on birth cohorts. We also carried out a large-scale simulation study to examine the impact of accelerated longitudinal designs compared to the single birth cohort on statistical inference. The following are key findings from the report. • The key difference between an accelerated longitudinal design and the single cohort study is that an accelerated longitudinal design incorporates multiple cohort groups at the onset of the study, enabling the duration of the study to be shortened. This investigation focuses on aspects of an accelerated longitudinal design compared to a single cohort design, particularly for following babies, children and youth into adulthood. We provide a comprehensive literature review and investigate aspects of the designs through a theoretical assessment and simulation study. • The theoretical assessment of the precision of basic longitudinal analyses under single, 2-cohort and 3-cohort designs shows that sample sizes can be compromised especially at the extreme ages of the cohorts. This problem can be addressed by refreshment samples that not only ensure representativity but also compensate for left- censored data. • The simulation study shows that the accelerated longitudinal designs permit analysis across a wider age span for a given duration of the study. The precision of parameter estimation for a multilevel growth curve model is similar for the alternative designs. If cohort effects are present in the study, it is important to include cohort main effects and their interactions in the model. For small sub-groups, the level of precision can be compromised due to small sample sizes. • The overall conclusion of this investigation regarding the suitability of an accelerated longitudinal design for the future of longitudinal study in the UK is positive. We recognize, nevertheless, that, in coming to a decision about the longitudinal design, careful consideration is needed of the different types of analysis undertaken by substantive researchers and their additional complexities under an accelerated longitudinal design. In addition, it may be important to consider other aspects of survey methodology impacted by an accelerated longitudinal design when coming to a decision, but these are outside the scope of this report. • From the review of international surveys of children and young people, it is clear that multiple cohort designs have been successfully implemented in different countries and the UK can learn from such experiences whilst considering specific needs, for example, boosting the 2016 'missed' cohort from the cancelled Life Style Study. The report has been added to the University of Manchester Cathie Marsh Institute Reports website: http://hummedia.manchester.ac.uk/institutes/cmist/archive-publications/reports/2019-01-Deliverable_Report_for_the_Economic_and_Social_Research_Council.pdf |
| Exploitation Route | Traditional single cohort studies have successfully been carried out in the UK, the most recent being the Millennium Cohort Study analysing a birth cohort from 2000. However, with increasing costs of implementing such studies, growing rates of nonresponse and attrition over the years and the length of time needed to obtain a significant number of observations for addressing key societal questions, there is a need to develop new ways of collecting longitudinal data over the life-course. This report provides the first steps in examining the potential for accelerated longitudinal designs with respect to their strengths and weaknesses. More generally, the report can inform other design changes in standard longitudinal studies, such as implementing a rotating panel design in Understanding Society. |
| Sectors | Healthcare,Government, Democracy and Justice |
| URL | http://hummedia.manchester.ac.uk/institutes/cmist/archive-publications/reports/2019-01-Deliverable_Report_for_the_Economic_and_Social_Research_Council.pdf |