C7 may have a point when he asks about the margin of error of individual data points. I am not sure if this concern has been adequately addressed by femr2, and if yes, if the error margin has been found to be small enough to retain confidence in the existence of an over-g period.
I simply don't know. I kinda remember that this was a big point of contention when he first presented his results here, but don't remember, or perhaps never saw, how this was resolved (also, I may mix this up with the "missing jolt" debate).
C7 is, however, wrong when he asserts that leaving out data points and going with averages over longer time intervals is a solution to the problem of margin of error. This approach smoothes out the signal as much as the error. That is, in effect, what NIST had done in their draft report: They took just two data points - beginning of north wall descent, and a moment just before it drops out of sight, and computed an average acceleration well below g. This method obviously would mask any period of at-g or over-g acceleration, if most of the fall was under-g. And the same happens if you elect to go with a coarser grid than what your original dataset allows. Chandler takes averages, just like NIST did originally, only for somewhat finer intervals, and his method, too is thus prone to mask real signals towards the extreme end of reality.
