4. Evaluation and quality control

The quality of experimental or observational data may vary widely, depending on the care taken by the scientist who did the research. Furthermore, most measurements depend on some form of calibration, which can change over the years. The risk has long been recognized, especially in the physical sciences, of assuming a piece of data taken from the literature is valid without further checking. A distinct methodology of data analysis and evaluation has evolved, leading to compilations of "evaluated data" that can be used with confidence by the general scientific community. The details of the methodology vary with the type of data but it usually includes a careful study of the way the measurement was made (as described by the author); application of various corrections needed because of changes in temperature scale, fundamental constants, and the like; and comparison with applicable theory. Ideally, this evaluation procedure is applied systematically to a large body of data, so that any discrepant numbers are more visible.

This approach to quality control is not so easily applied in the geosciences and biosciences because of the different nature of the data, as already discussed. The most important consideration is to establish quality control before the experiment or observation is made. Thus the calibration of the instruments should be carefully documented and a valid statistical design established. Nevertheless, an independent peer review after the results are published often turns up errors and inconsistencies.