|
Confounding by group |
|||||||||
| Group 0 |
Group 1 |
Crude |
|||||||
|
|
|
|
|||||||
| expose |
unexposed |
sum |
expose |
unexpose |
sum |
expose |
unexpose |
sum |
|
|
|
|||||||||
| case |
16 |
12 |
28 |
48 |
24 |
72 |
64 |
36 |
100 |
| noncase |
24 |
48 |
72 |
12 |
16 |
28 |
36 |
64 |
100 |
| total |
40 |
60 |
100 |
60 |
40 |
100 |
100 |
100 |
200 |
| risk |
0.40 |
0.20 |
0.80 |
0.60 |
0.64 |
0.36 |
|||
| RD |
0.20 |
0.20 |
0.28 |
||||||
| Xi |
0.40 |
0.60 |
|||||||
| Yi |
0.28 |
0.72 |
|||||||
| ni |
100 |
100 |
|||||||
|
Since background risks and exposure distributions differ between the two groups, the crude individual-level estimate of the risk difference is confounded. The ecologic estimate of the risk difference, (Y1-Y0)/(X1-X0) = 2.2, is much more confounded. RD = risk difference; Xi and Yi are the average exposure and average risk in group i; ni = size of group i. | |||||||||
Webster Environmental Health 2007 6:17 doi:10.1186/1476-069X-6-17 |
|||||||||