Table 4.
Correct | Misclassified | |||||
expose | unexpose | sum | expose | unexpose | sum | |
Case | a i | b i | ai+bi | sai + (1-t)bi | (1-s)ai+tbi | ai+bi |
noncase | c i | d i | ci+di | sci + (1-t)di | (1-s)ci+tdi | ci+di |
Total | ai+ci | bi+di | n i | s(ai+ci) + (1-t)(bi+di) | (1-s)(ai+ci) + t(bi+di) | n i |
X i | (ai+ci)/ni | s(ai+ci)/ni + (1-t)(bi+di)/ni | ||||
Y i | (ai+bi)/ni | (ai+bi)/ni |
Assume that the proportion of people misclassified does not depend on disease status. Sensitivity s is the fraction of exposed people classified as exposed; specificity t is the fraction of unexposed people classified as unexposed. On the group level, the average risk remains the same but the average exposure changes. Xi and Yi are the average exposure and average risk in the group; ni = size of the group.