Table 3.
Group 0 | Group 1 | Crude | |||||||
expose | unexpose | sum | expose | unexpose | sum | expose | unexpose | sum | |
case | 20 | 10 | 30 | 48 | 12 | 60 | 68 | 22 | 90 |
noncase | 80 | 90 | 170 | 32 | 108 | 140 | 112 | 198 | 310 |
total | 100 | 100 | 200 | 80 | 120 | 200 | 180 | 220 | 400 |
risk | 0.2 | 0.1 | 0.6 | 0.1 | 0.378 | 0.1 | |||
RD | 0.1 | 0.5 | 0.278 | ||||||
X i | 0.5 | 0.4 | |||||||
Y i | 0.15 | 0.3 | |||||||
n i | 200 | 200 |
The background risks in each group are the same but the risk differences vary. The crude individual-level risk difference lies between the two extremes, as it must when exposure is binary. The ecologic estimate of the risk difference, (Y1-Y0)/(X1-X0)= -1.5, is extremely biased. RD = risk difference; Xi and Yi are the average exposure and average risk in group i; ni = size of group i.