soulor said:The combination of measurement uncertainty is not always straightforward (correct me if I am wrong, please), and I feel the need for some clarity on how the total error is being calculated here. Clearly if there are measurement errors as large as 11% for an estimate of power to weight then these may make the whole effort statistically futile (as Coggan argues), but if the different errors are simply down to different methods of error calculation then I am left really confused and rather unenlightened. In my field there have been some very fruitful discussions about measurement error which have helped identify the limits of particular measurement tools. What is interesting is that Tucker and Dugas do state that their error estimates are based on comparison with mechanical measurements. Many apologies if this has been dealt with elsewhere in detail or I am myself just suffering from confusion about error and uncertainty. BTW I hope Tucker and Dugas have not given up on this forum entirely...
As I indicated previously, I used a standard propogation-of-error approach, assuming a 2% error in estimating the relevant variables.
In contrast, Le Breton not only started from different assumptions, but also simply assumed that any errors were additive, not multiplicative.
Finally, I do not know the exact basis for Tucker's assertions, as I have been unable to locate the paper to which he alluded.
Obviously, though, the last approach described would seem to be the preferable one to take, especially since comparisons were made to direct measurements of power. However, even if the true confidence interval is +/- ~0.4 W/kg as he/they implied versus the +/- ~0.6 W/kg suggested by a straight propogation-of-error analysis, I still don't think the overall approach has merit, especially given the uncertainty on the physiological supply side of things.