Figure 5
From: Bias magnification in ecologic studies: a methodological investigation
Confounding by group on the individual and group level. A, B, C) Suppose average exposures are the same, but the difference between the background risks (q i ) decreases. Confounding by group decreases on both the individual and group levels with constant proportionality factor M. D, E, F) Suppose background risks (q i ) are the same, but the difference between the average exposures decreases. Confounding by group decreases on the individual level, but increases on the ecologic level because of the large increase in M.