From samples and various trials, estimates can be made of statistics for distributions and the effects of different factors. If sampling and trials are properly planned, it should be possible to calculate an interval (confidence interval) which for a given probability covers the right value of a statistic or effect, as well as perform hypothesis testing with given probabilities of making errors of the first kind (reject a true hypothesis) and the second kind (accept a false hypothesis).
Examples of hypotheses which can be tested are:
- the mean value of a population has a certain value;
- the mean values of two populations are equal;
- observed values come from a population with a given distribution;
- two variables are independent (for example no time dependence, i e no trends in the results);
- a measurement value is an outlier, that is it cannot be considered to belong to the same distribution as the other values