Check randomness of sample data

When calculating the parameters of a probability law from a sample, it is essential to make sure that the data available come from a truly random sample. It is not uncommon for systematic effects to mix with random phenomena. For example, when repeating measurements on a steel part to assess dispersion (repeatability), the part may undergo expansion during the course of the measurements. This dilation adds to the purely random effects, and calculating a standard deviation on a series of data contaminated by such a phenomenon distorts the estimate.

This type of situation can often be detected by a simple graphical analysis (see figure Time representation of a series of values).