Cost-Justifying Usability, 2nd Edition

18.4.3 Do We Need Hypotheses?

A steady undercurrent claiming that we do not actually need the apparatus of

null hypothesis formulation is present in statistical thinking. What we are most

concerned with, this argument states, is not the blanket acceptance or rejection

of a hypothesis, but an idea of where the true population parameters lie given

what we know about the sample. This usually resolves into the estimation of what

are known as the “confidence intervals of the mean (or median).” The 95% con-

fidence intervals are the ones usually used. The confidence interval of the mean

does not actually make any statements about where the real population mean ...