Using Uncertainty Analysis in the Planning of an Experiment

Abstract

A simple example using convection heat transfer is used to illustrate the use of uncertainty analysis in PLANNING experiments. Major points made are: (i) the choice of test and data-reduction procedure can have important impact on the accuracy of the results, with one procedure better for some conditions and the other better in other ranges; (ii) it is important to specify carefully the level of replication (what is held constant and what varied in a given test), since otherwise an inappropriate value of uncertainty may be generated; (iii) reliable means for cross-checking and/or externally validating the results of an experiment are necessary if predicted uncertainties are to be confirmed; (iv) in experiments where data are reduced by computer, uncertainty analysis can be done by sequential perturbation, using the main data-reduction program itself.