As biomechanists, we try to optimize the movement of athletes, patients, and others. We may be trying to maximize performance, minimize the risk of injury, or simply returning someone to “normal.” A term often used in biomechanics is “kinetic chain.” But what is a kinetic chain? Considering that this is a fundamental concept of biomechanics, there is a surprising lack of clarity on its definition. In his seminal textbook, Arthur Steindler defined a kinetic chain vaguely as “a combination of several successively arranged joints constituting a complex motor unit.”^{1} More recently, the kinetic chain has been explained as “an interaction of body segments”^{2}, a “transfer of forces and motion”,^{3} and a “sequential transfer of energy.”^{4}

So, here is my question for discussion. What is the most meaningful way to quantify the kinetic chain from our biomechanical data? That is, as biomechanists, do we ideally want to quantify the kinetic chain as the magnitudes and sequential timing of:

I welcome your expertise. Please vote and/or (more importantly) post your thoughts.

]]>So, here is my question for discussion. What is the most meaningful way to quantify the kinetic chain from our biomechanical data? That is, as biomechanists, do we ideally want to quantify the kinetic chain as the magnitudes and sequential timing of:

- Joint angles?
- Joint angular velocities?
- Joint torques?
- Segmental energy?
- Muscle firing patterns?
- Something else?

I welcome your expertise. Please vote and/or (more importantly) post your thoughts.

We might envision a repeated-measures design with multiple sprinting trials for each condition (no fatigue assumed). Correlated measurements would be expected, so we need to employ a repeated-measures test. The expectation is that multiple trials helps us arrive at a value that is closer to 'reality'.

However, what if each sprinting trial involves multiple steps (peak forces following contact, for example)? The measurements corresponding with each step would also be correlated within each trial. Since there is an expected improvement in statistical power when we employ a repeated-measures design, how can we take advantage of multiple steps?

Another example might be calculating average values over of some measure the entire gait cycle from running trials collected for 1 minute. At 90 strides/180 steps per minute, that is a lot of measurements. It would seem that we would fail to benefit from statistical power considerations if we simply put one single average value from that 1 minute trial into the hypothesis test calculations.

Conceptually or otherwise, is there such a thing as a within-*trial *factor nested inside a within-*subject* factor?

]]>However, what if each sprinting trial involves multiple steps (peak forces following contact, for example)? The measurements corresponding with each step would also be correlated within each trial. Since there is an expected improvement in statistical power when we employ a repeated-measures design, how can we take advantage of multiple steps?

Another example might be calculating average values over of some measure the entire gait cycle from running trials collected for 1 minute. At 90 strides/180 steps per minute, that is a lot of measurements. It would seem that we would fail to benefit from statistical power considerations if we simply put one single average value from that 1 minute trial into the hypothesis test calculations.

Conceptually or otherwise, is there such a thing as a within-