Browsing by Author "Adeboye, Promise Anjolaoluwa"

Now showing 1 - 1 of 1
  • Thumbnail Image
    ItemOpen Access
    Cross-sectionally Exploring Proxy Usage for Historical Data Analysis: A Proof of Concept
    (2024-11-01) Adeboye, Promise Anjolaoluwa; McDonald, Sheila; Suzanne, Tough; Patten, Scott; Holodinsky, Jessalyn
    Introduction: Longitudinal cohorts routinely collect information that encompass many different aspects of the life course, but their analysis is often restricted to validated scales. Such validated scales are often routinely shifting to reflect contemporary understandings of important concepts. Such routine changes often mean that a present-day scale measuring a concept may be incomparable to a different scale measuring the same concept in previous time-points. This restriction greatly hampers the strength of longitudinal cohorts in being able to make comparisons across time points. In this problem lies an opportunity to create a method of proxy development such that the proxy would be impervious to scale evolution over time. Flourishing is a common, well-studied concept that is a good candidate for this proof of concept. This thesis explores the use of proxy measurements as an optimization procedure for longitudinal cohorts going forward. Methods: As a proof of concept, two methods were explored for proxy creation using the test concept of flourishing. Using a bootstrapped linear regression approach, the proxy models were evaluated on performance metrics. Discussion: Although the proxy models performed similarly, the interpretation of results differed per model. Through the development of a flourishing proxy based on available data, it is possible to quantify historical flourishing, such that associations can be made between that and other concepts. Due to the context of the model derivation, the coefficients of either models cannot be extrapolated to other cohorts, but the coefficients of the knowledge-driven model can be furthur validated temporally within the derivation cohort. The included variables from both proxies contribute to the conceptual understanding of flourishing, but only in this specific cohort. Conclusion: Proxy derivation methods have the capacity to increase the value-add of historically collected data. However, more evaluations and validations of both the coefficients and the derivation method are needed before this can be used as a finalized procedure.