Building confidence intervals around the obtained value of a sample
Dominique M. Francois-Bongarcon, PhD Agoratek International Consultants Inc., North Vancouver, Canada. E-mail: [email protected]
Common practice in sampling for the TOS erudite consists of using the sampling variance obtained from Gy’s numerical theory to build confidence intervals around the true sample value. This is usually done to characterise the ‘precision’ of the sample, and, by centring that interval on the sampled value, one states for instance that “the true value has 95% chances of being between values x and y”, those two values usually being centred on the sampled value”. The somewhat naïve rationale behind this practice is reviewed in some details and criticised. It is suggested the confidence interval of real interest to the user of the sampled value, is more difficult to define and more delicate and indirect to build. Some methods for doing so are examined and a methodology is recommended.