Skip to main content

Advertisement

Table 1 A hypothetical population with positive/negative confounding U

From: Detecting and correcting the bias of unmeasured factors using perturbation analysis: a data-mining approach

Level of U (i) Population number (m i ) Positive confounding Negative confounding
Exposure prevalence (p i ) Disease risk among the unexposed ( r i u ) Relative risk ( r i e / r i u ) Exposure prevalence (p i ) Disease risk among the unexposed ( r i u ) Relative risk ( r i e / r i u )
1 2,500 0.76 0.6667 1.2632 0.24 0.4737 1.7593
2 2,500 0.60 0.4000 1.1667 0.40 0.3333 2.4000
3 2,500 0.40 0.3333 1.2000 0.60 0.2000 2.3333
4 2,500 0.24 0.1579 1.0556 0.76 0.1667 1.8947
Total 10,000 Crude relative risk (crude RR) = 1.75 Crude relative risk (crude RR) = 1.53
Standardized relative risk (SRR) = 1.20 Standardized relative risk (SRR) = 2.06