TY - JOUR
T1 - Erratum
T2 - Errata to “Modelling muscle recovery from a fatigued state in isometric contractions for the ankle joint” (Journal of Biomechanics (2020) 100, (S002192902030004X), (10.1016/j.jbiomech.2020.109601))
AU - Rakshit, Ritwik
AU - Yang, James
N1 - Publisher Copyright:
© 2020
PY - 2020/5/22
Y1 - 2020/5/22
N2 - The authors wish to draw your attention to an incorrectly plotted error curve for the superelliptic model in the published form. The corrected Fig. 2 is as shown below: [Figure presented] While the conclusions remain unaffected as they were based on analyses that still hold true, two portions of the discussion addressing the shape of the superelliptic model's error curve have been rewritten to reflect the properties of the corrected shape. Results and discussion, paragraph 3: “The step, linear and MSR models have a minimum error corresponding to a rest-recovery multiplier (rTL=0 or rrest) of 15– 17, while the superellipse predicts minimum error at rrest = 4 as seen in Column 4 of Table 2. This difference between the values of rrest for the first three models is unlikely to be meaningful given the inherent noise in fatigue data, and all three approaches predict essentially equivalent errors using virtually equivalent rest recovery parameters. The superelliptical (n = 8) curve requires the lowest rrest = 4, but one-way ANOVA performed on the errors generated by the step and the superellipse-8 indicates that there is no significant difference between the models at their respective koptimum values (p = 0.49).” This paragraph should be replaced by the following one: All models tested have a minimum error corresponding to a rest recovery multiplier of 15–17 as seen in Column 4 of Table 2, but the difference between the values of rrest for these models is unlikely to be meaningful given the inherent noise in fatigue data, and all approaches predict essentially equivalent errors using virtually equivalent rest recovery parameters. Indeed, one-way ANOVA performed on the errors generated by the superelliptical and the step curves indicates that there is no significant difference between the models (p > 0.99) at their respective koptimum values. Results and discussion, paragraph 5: “Whereas the linear and the step models have nearly the same modest sensitivity, the MSR and the superelliptic models have a somewhat smaller sensitivity throughout the first 100 values of k, implying that selecting a common value for koptimum will be easier. The errors for both models range almost monotonically from 8.27% to 16.12% as seen in Fig. 2. The curves depict the variation of mean RMS error between experimental and model data with k for each of the four models: the original 3CC-r with a step function, the superelliptic, the MSR, and the linear models.” This paragraph should be replaced by the following one: The superelliptic, linear, and step models all have nearly the same modest sensitivity to the optimization parameter k around their respective values of koptimum, while the MSR has a somewhat smaller sensitivity throughout the first 100 values of k as seen in Fig. 2. While this implies that choosing a common k-value should be easier in the case of the MSR model, it must be noted that being able to model fatigue for all joints with a common k is a consideration determinedly secondary to model accuracy.
AB - The authors wish to draw your attention to an incorrectly plotted error curve for the superelliptic model in the published form. The corrected Fig. 2 is as shown below: [Figure presented] While the conclusions remain unaffected as they were based on analyses that still hold true, two portions of the discussion addressing the shape of the superelliptic model's error curve have been rewritten to reflect the properties of the corrected shape. Results and discussion, paragraph 3: “The step, linear and MSR models have a minimum error corresponding to a rest-recovery multiplier (rTL=0 or rrest) of 15– 17, while the superellipse predicts minimum error at rrest = 4 as seen in Column 4 of Table 2. This difference between the values of rrest for the first three models is unlikely to be meaningful given the inherent noise in fatigue data, and all three approaches predict essentially equivalent errors using virtually equivalent rest recovery parameters. The superelliptical (n = 8) curve requires the lowest rrest = 4, but one-way ANOVA performed on the errors generated by the step and the superellipse-8 indicates that there is no significant difference between the models at their respective koptimum values (p = 0.49).” This paragraph should be replaced by the following one: All models tested have a minimum error corresponding to a rest recovery multiplier of 15–17 as seen in Column 4 of Table 2, but the difference between the values of rrest for these models is unlikely to be meaningful given the inherent noise in fatigue data, and all approaches predict essentially equivalent errors using virtually equivalent rest recovery parameters. Indeed, one-way ANOVA performed on the errors generated by the superelliptical and the step curves indicates that there is no significant difference between the models (p > 0.99) at their respective koptimum values. Results and discussion, paragraph 5: “Whereas the linear and the step models have nearly the same modest sensitivity, the MSR and the superelliptic models have a somewhat smaller sensitivity throughout the first 100 values of k, implying that selecting a common value for koptimum will be easier. The errors for both models range almost monotonically from 8.27% to 16.12% as seen in Fig. 2. The curves depict the variation of mean RMS error between experimental and model data with k for each of the four models: the original 3CC-r with a step function, the superelliptic, the MSR, and the linear models.” This paragraph should be replaced by the following one: The superelliptic, linear, and step models all have nearly the same modest sensitivity to the optimization parameter k around their respective values of koptimum, while the MSR has a somewhat smaller sensitivity throughout the first 100 values of k as seen in Fig. 2. While this implies that choosing a common k-value should be easier in the case of the MSR model, it must be noted that being able to model fatigue for all joints with a common k is a consideration determinedly secondary to model accuracy.
UR - http://www.scopus.com/inward/record.url?scp=85084048199&partnerID=8YFLogxK
U2 - 10.1016/j.jbiomech.2020.109802
DO - 10.1016/j.jbiomech.2020.109802
M3 - Comment/debate
C2 - 32362406
AN - SCOPUS:85084048199
VL - 105
JO - Journal of Biomechanics
JF - Journal of Biomechanics
SN - 0021-9290
M1 - 109802
ER -