A comparison of risk in two groups may be distorted by many confounding factors, some causing overestimation and others causing underestimation of risk.