Please use this identifier to cite or link to this item: doi:10.22028/D291-36013
Title: Dwarfs on the Shoulders of Giants: Bayesian Analysis With Informative Priors in Elite Sports Research and Decision Making
Author(s): Hecksteden, Anne
Forster, Sabrina
Egger, Florian
Buder, Felix
Kellner, Ralf
Meyer, Tim
Language: English
Title: Frontiers in Sports and Active Living
Volume: 4
Publisher/Platform: Frontiers
Year of Publication: 2022
Free key words: Bayesian statistics
individual response
cold-water immersion
methodology
replicate crossover
DDC notations: 570 Life sciences, biology
Publikation type: Journal Article
Abstract: While sample sizes in elite sports are necessarily small, so are the effects that may be relevant. This conundrum is complicated by an understandable reluctance of athletes to comply with extensive study requirements. In Bayesian analyses, pre-existing knowledge (e.g., from sub-elite trials) can be formally included to supplement scarce data. Moreover, some design specifics for small sample research extend to the extreme case of a single subject. This provides the basis for actionable feedback (e.g., about individual responses) thereby incentivising participation. As a proof-of-concept, we conducted a replicated cross-over trial on the effect of cold-water immersion (CWI) on sprint performance recovery in soccer players. Times for 30 m linear sprint and the initial 5 m section, respectively, were measured by light gates before and 24 h after induction of fatigue. Data were analysed by Bayesian and by standard frequentist methods. Informative priors are based on a published metaanalysis. Seven players completed the trial. Sprint performance was 4.156 ± 0.193 s for 30 m linear sprint and 0.978 ± 0.064 s for the initial 5 m section. CWI improved recovery of sprint time for the initial 5 m section (difference to control: -0.060 ± 0.060 s, p = 0.004) but not for the full 30 m sprint (0.002 ± 0.115 s, p = 0.959), with general agreement between Bayesian and frequentist interval estimates. On the individual level, relevant differences between analytical approaches were present for most players. Changes in the two performance measures are correlated (p = 0.009) with a fairly good reproducibility of individual response patterns. Bayesian analyses with informative priors may be a practicable and meaningful option particularly for very small samples and when the analytical aim is decision making (use / don't use in the specific setting) rather than generalizable inference.
DOI of the first publication: 10.3389/fspor.2022.793603
Link to this record: urn:nbn:de:bsz:291--ds-360131
hdl:20.500.11880/32814
http://dx.doi.org/10.22028/D291-36013
ISSN: 2624-9367
Date of registration: 20-Apr-2022
Description of the related object: Supplementary Material
Related object: https://www.frontiersin.org/articles/10.3389/fspor.2022.793603/full#supplementary-material
Faculty: HW - Fakultät für Empirische Humanwissenschaften und Wirtschaftswissenschaft
M - Medizinische Fakultät
Department: HW - Wirtschaftswissenschaft
M - Sport- und Präventivmedizin
Professorship: HW - Prof. Dr. Ralf Kellner
M - Prof. Dr. Tim Meyer
Collections:SciDok - Der Wissenschaftsserver der Universität des Saarlandes

Files for this record:
File Description SizeFormat 
fspor-04-793603.pdfPublikationPDF874,5 kBAdobe PDFView/Open


This item is licensed under a Creative Commons License Creative Commons