In experimental science, answering a question for an entire population can be complicated because of population size. Random sampling of the population allows to estimate the answers to these questions. We will see how to evaluate the confidence in these estimates in the simple framework where only one question is asked, as well as the treatments to be made in the framework of several questions.