EN
In the sampling surveys we have got three important values related in the strict mathematical way: the sampling error (usually 4%), the probability that sampling device does not charge the results with an error bigger than 3% (it is called confidence level and usually it is 0,95) and the sample size (about 1067 people). The article explains the correlation stressing that the 3% statistic error mentioned in the official notes informing on public opinion surveys is not the only one. Drawing on the Gallup Institute experience the article stresses the role of the additional information on the individuals participating in the survey in reducing other (non-sampling) errors.