Search

Confounding can make an association seem bigger when the true effect is smaller or vice-versa and it can also make it appear negative when it may actually be positive. In short, both the direction and the magnitude of an association are dependent on confounding. Therefore, understanding and adjusting for confounding in epidemiological research is central to addressing whether an observed association is causal or not. Moreover, unmeasured confounding in observational studies can give rise to biased estimates. Several techniques have been developed to account for bias and conducting sensitivity analysis. Using an hypothetical example this paper illustrates application of simple methods for conducting sensitivity analysis for unmeasured confounder(s). [ABSTRACT FROM AUTHOR]