Supplemental Table 1. The impact of cohort size on error due to stochastically differential sensitivity. All scenarios below feature perfect specificity and baseline risk equal to (R0) set to 0.01. Cohort sizes of 10,000 (i.e., our main simulation) and 1,000,000 are compared with respect to central tendency and dispersion of DeltaSe (sensitivity in the exposed minus sensitivity in the unexposed) and the observed bias factor.
|
Scenario parameters |
Cohort size = 10,000 |
Cohort size = 1,000,000 |
||||
|
Sensitivity |
Exposure prevalence (PE) |
True risk ratio (RRT) |
DeltaSe, median (range) |
Bias factor (RRM/RRS), geometric mean (range) |
DeltaSe, median (range) |
Bias factor (RRM/RRS), geometric mean (range) |
|
0.5 |
0.1 |
1.0 |
0 (-0.53, 0.61) |
0.947 (0.141, 2.594) |
0 (-0.06, 0.07) |
1.000 (0.877, 1.132) |
|
2.0 |
0 (-0.45, 0.51) |
0.977 (0.128, 2.444) |
0 (-0.05, 0.04) |
1.000 (0.904, 1.088) |
||
|
5.0 |
0 (-0.34, 0.34) |
0.995 (0.406, 1.982) |
0 (-0.03, 0.03) |
1.000 (0.932, 1.069) |
||
|
0.5 |
1.0 |
0 (-0.39, 0.40) |
0.998 (0.457, 2.476) |
0 (-0.04, 0.04) |
1.000 (0.933, 1.074) |
|
|
2.0 |
0 (-0.37, 0.31) |
1.006 (0.475, 2.098) |
0 (-0.03, 0.03) |
1.000 (0.934, 1.065) |
||
|
5.0 |
0 (-0.31, 0.29) |
1.009 (0.597, 2.419) |
0 (-0.03, 0.03) |
1.000 (0.935, 1.057) |
||
|
0.8 |
0.1 |
1.0 |
0 (-0.72, 0.35) |
0.985 (0.178, 1.542) |
0 (-0.05, 0.05) |
1.000 (0.932, 1.062) |
|
2.0 |
0 (-0.42, 0.35) |
0.994 (0.495, 1.538) |
0 (-0.04, 0.04) |
1.000 (0.952, 1.046) |
||
|
5.0 |
0 (-0.28, 0.24) |
0.999 (0.658, 1.336) |
0 (-0.03, 0.03) |
1.000 (0.968, 1.034) |
||
|
0.5 |
1.0 |
0 (-0.35, 0.30) |
1.000 (0.613, 1.473) |
0 (-0.03, 0.03) |
1.000 (0.961, 1.036) |
|
|
2.0 |
0 (-0.29, 0.27) |
1.000 (0.703, 1.484) |
0 (-0.03, 0.03) |
1.000 (0.969, 1.035) |
||
|
5.0 |
0 (-0.24, 0.24) |
1.001 (0.755, 1.408) |
0 (-0.02, 0.02) |
1.000 (0.970, 1.028) |
||