All marker positional data were filtered

using the same filter level reported by Brice et al.3 Positional data were then used in conjunction with direction cosines to determine the three-dimensional coordinate data for the centre of the hammer’s head. These positional data were used to calculate hammer linear velocity (calculated speed) and cable force (calculated force).3 All calculated and measured force data were normalised for hammer weight to account for the fact that males use a heavier hammer than females. Two regression models were developed that allowed speed to be predicted from measured find more force data (predicted speed). The calculated speed data and calculated force data were used to develop these regression models. All calculated speed data used in the regression model development were squared due to the mechanical relationship that exists between centripetal force and linear velocity squared (Equation (1)). The first regression model was derived

from the square of the calculated speed and the calculated CFTR activator force (non-shifted regression). While the second model was derived from the square of the calculated speed and a time shifted calculated force (shifted regression). The shifted regression model was developed because earlier work showed a phase lag between speed and cable force3 and it was thought that accounting for the phase lag in the model development may lead to a model that would produce speed data that were more accurate. As the magnitude of this phase lag varies

depending on turn number, throw, and athlete, it is not possible to Ketanserin apply the same time shift to every throw. It was therefore decided to time shift the calculated force such that for each throw the final peaks in the calculated force and calculated speed coincided. This time shift was applied to ascertain if removal of the phase lag resulted in a more accurate regression. As only the final peaks were aligned, there was no change in the frequency of the force data. The calculated speed and calculated force data used to calculate the shifted regression were also trimmed as the final peak in the calculated force data occurred prior to release whereas the final peak in speed occurred at release. The calculated force data were trimmed so that the final peak was the final data point and the calculated speed data were trimmed by the same amount at the start. This was done so that both data sets were the same size. A shifted and non-shifted regression equation was developed for each of the participant’s 10 throws and all data points of each throw were used to develop these equations. The MATLAB software suite (The Mathworks, Natick, MA, USA) was used to determine the regression equations and the y intercepts for both were also forced through (0,0) since Equation (1) predicts zero speed for zero force. Averages of the gradients of the two linear regression equations were determined for the cohort.