What correlation?
Problem: The correlation between two tests is influenced by the base rate, the proportion of patients who have the disease.
- Therefore there is no “correlation between two tests” characteristic of all situations. Situations with different disease prevalences would have different correlations.
Solution: If we use two correlations
- between false negatives, for patients with the disease
- between false positives, for patients without the disease
then it is possible to vary the baserate without affecting the correlations.