ABSTRACT
Hydraulic experiments are performed with different aims, and hence the appropriate data analysis will differ. In a scale-model experiment, where a single value is to be obtained, then a simple uncertainty analysis may suffice. Two general principles are stressed throughout: physics supercedes statistics, or because statistical significance does not imply physical or engineering significance, the latter is cautiously given precedence if a conflict between the two is found, and the simpler or the more parsimonious model is preferred, provided no well-founded physical principle is violated. Statistics is applied to a quantity thought to vary randomly, that is, in an 'unpredictable' manner, to some degree, and hence is a random variable. Hydraulic experiments are performed in order to characterize the variation of a flow or transport variable in space or time or as other variables change. Bayesian approaches to time series and specifically spectral analysis are available, and that bootstrapping can be applied to obtain confidence intervals for power spectral estimates.
