Background: In order to find out truth about the target population size of the study sample has to be large enough to contain the characteristic (variable) we are seeking with variability sufficient to reliably reflect its variability in the population. Objective: The aim of this article was to explain logic of sample size calculation in comparative studies, and shed some light on key assumptions of the calculation. Methods: This article is a review of methodology used for estimating appropriate size of a study sample. Results: True difference in target parameter among the populations that are studied, and its variability (usually expressed as standard deviation from the mean) could not be changed according to our preferences; also maximum acceptable levels of probability if type one and type two errors cannot be further increased without compromising ability of the study to give us reliable information about the populations. What we can change is number of patients within the study groups, which if increased, will decrease variability of the results, and make distribution of the difference between the groups (if the study is hypothetically repeated many times) around true value of difference between the populations more narrow. Through narrowing of the distributions we will decrease number of cases when the difference among the group (type one error) or lack of difference (type two error) happens by chance, i.e. put probabilities of these errors below limits of acceptability. Conclusion: Careful calculation of sample size is necessary to minimize probability of type one and type two errors and therefore obtain reliable answer to a research question.
Key words: sample size; statistical power; type one error; type two error.
|
