No matter how carefully a sample is drawn, there is no way that it can precisely reproduce all the details of the population it is taken from. When several samples are drawn from the same population, it becomes evident that the measured values for one and the same characteristic differ from one another in the different subgroups.
But which of the values measured is the right one or, in other words, how precisely do these indicators approximate the actual value?
The quality of a statement made on the basis of a sample can be narrowed down with mathematical precision: It primarily depends on the actual precision required - the significance level, the size of the sample and, depending upon the indicator analysed, the distribution of the indicator or the measured value.
Conversely, it is also true that the minimum size of a sample necessary for the desired significance level can be unambiguously defined.
Different calculation methods are applied depending on the underlying question and whether the sample is compared with the base population or two samples are compared with each other, and whether proportional values or means are to be interpreted.