Epidemiologic research yields an estimate of an association or an effect. This estimate is obtained from a sample representing the population. While the true values in the population are not known, it is assumed that the estimated values are similar to them. The bias, the unknown deviation from [Page 129]the true value, determines the internal validity of a conclusion. The source of potential bias or confounding is sometimes obvious but more commonly unknown. In practice, researchers are expected to avoid known sources of bias in the design, control for potential bias in the analysis, and replicate their results to increase the internal validity of their studies.