Last update:
06-Jun-2005

Measurement error refers to deviations between the estimated and the true value of a (biological) variable, which are not caused by mistakes. In epidemiological studies, errors are classified into random and systematic and the three categories of systematic error are selection bias, information bias and confounding. Selection biases result from procedures used to select the study population and from factors that influence study participation or participation in a specific group of the study. In cohort studies, these errors occur when participation is associated with the outcome under study and in case-control studies when participation is associated with the determinant. Information biases are introduced in a study when collected information regarding the category of the determinant, the outcome, or both is distorted. When the characteristic under study is qualitative and error leads some persons to be included in a wrong category, the bias is a misclassification. Finally, confounders are the external determinants of the frequency of the disease that are associated with the determinant under study and are not intermediate steps in the causal mechanism between the determinant and the disease. As opposed to selection and information biases, confounders can be controlled during both study design and data analysis.

Key words: Bias, Confounder, Information bias, Selection bias, Validity.