On correct computation of confidence intervals for kinetic parameters

摘要

Two papers have been recently published in PhysiologyReports (Goulding et al. 2018a,b) that compare thekinetic parameters of three physiological signals duringtransitions in exercise intensity. The authors, using acommercial statistical software to apply the nonlinearregression technique, reported among their results also“the 95% confidence intervals (CIs) for the derivedparameter estimates”. The investigated signals (oxygenuptake, heart rate, and NIRS data) were acquired withdifferent time resolutions, that is breath-by-breath, at 1 s(1 Hz) and 0.5 s (2 Hz), respectively. Before running thenonlinear regression, the oxygen uptake data only underwentan interpolation procedure to produce second-bysecondvalues (i.e. at 1 Hz). It should be noted, however,that this procedure was not even supported by the resultsreported by Benson et al. (2017), although it was suggestedin their abstract (Francescato et al. 2017). In fact,this procedure does not add new information to the data,rather it reiterates the already available information in thenewly introduced data points, invalidating the CIsobtained by the calculations (Francescato et al. 2015).