Abstract
The Bayesian paradigm holds that previous knowledge or expertise (“prior”) and data-driven measurements (“likelihoods”) can produce an outcome (“posterior”) which can, in turn, be used as the prior in a new iteration of analysis and learning. In a wider project, we are undertaking to measure hamstrings strength, physical activity level, and ultrasound (US) and MRI characteristics of more than 200 participants ages 18+. PURPOSE: Here we aim to develop a model that will predict eccentric hamstrings strength as measured by Nordbord based on variables available in both surveys (IPAQ) and US for use as a starting point for our analysis of future participants. METHODS: Eleven participants (age 24.2 ± 4.8 y, ht 175.4 ± 11.7 cm, wt 73.2 ± 15.5 kg, F = 5) will responded to the IPAQ short form, had hamstrings size measured via US, and performed an eccentric Nordic curl on Nordbord. With this data, we tested eight iterations of a Bayesian formula for goodness of fit and overfitting. We examined the predictive effects of variables of interest on total force generated during an eccentric hamstring curl. We compared various combinations of these variables, and selected the model with the lowest Leave-One-Out Cross-Validation (LOO) score as the optimal model. We required enough independent samples for reliable parameter estimates (bulk effective sample sizes ≥ 3000) with good convergence (R-hat = 1.00). RESULTS: After ensuring our requirements were met, we examined eight models, comparing LOO scores. The initial model predicted Total Force with Age, Sex, Height, Weight, Average Echo Intensity, Truncated Volume, and Total weekly physical activity hours (LOO: 154.6 ± 9.4 SE). The best model predicted TotalForce with Sex, Height, Weight, Average Echo Intensity, Truncated Volume, and Moderate Intensity Physical Activity Hours (LOO: 148.6 ± 7.4). This model was based on all variables beginning with an uninformed normal prior of 0 ± 10. CONCLUSION: The Bayesian paradigm provides useful ways to learn from pilot data, which can help create more informative priors to use in the next iteration of analysis. We expect age to eventually be a more predictive variable because these participants were young, age is known to impact muscle strength and quality.
Recommended Citation
Hinkle, Lee J; Mead, James D.; Bates, Carson; and Johnson, Aaron W.
(2024)
"Determining Model Fit for Eccentric Hamstrings Strength: A Bayesian Approach,"
International Journal of Exercise Science: Conference Proceedings: Vol. 14:
Iss.
4, Article 47.
Available at:
https://digitalcommons.wku.edu/ijesab/vol14/iss4/47
Included in
Health and Physical Education Commons, Medical Education Commons, Sports Sciences Commons