Impact of stochastically differential sensitivity and specificity on bias in the risk ratio
Plots depict coefficients from multivariable linear regression of log10 bias factor on quantiles of 𝚫Se and 𝚫Sp (reference category=0)
Exposure prevalence = 0.1 | Risk in unexposed = 0.01
True RR = 1
Se = 0.8 
Se = 0.5 
True RR = 1.5
Se = 0.8 
Se = 0.5 
True RR = 2
Se = 0.8 
Se = 0.5 
True RR = 5
Se = 0.8
Se = 0.5
Exposure prevalence = 0.1 | Risk in unexposed = 0.10
True RR = 1
Se = 0.8
Se = 0.5
True RR = 1.5
Se = 0.8
Se = 0.5 
True RR = 2
Se = 0.8 
Se = 0.5 
True RR = 5
Se = 0.8 
Se = 0.5
Exposure prevalence = 0.5 | Risk in unexposed = 0.01
True RR = 1
Se = 0.8 
Se = 0.5 
True RR = 1.5
Se = 0.8 
Se = 0.5
True RR = 2
Se = 0.8
Se = 0.5
True RR = 5
Se = 0.8
Se = 0.5 
Se = 0.5
Se = 0.5
Se = 0.5
