INSTRUMENTAL VARIABLES IN THE EVALUATION OF DIAGNOSTIC-TEST PROCEDURES WHEN THE TRUE DISEASE STATE IS UNKNOWN

We explore the estimation of sensitivity and specificity of diagnostic tests when the true disease state is unknown. Instrumental variables which subdivide the patient population are used. A logistic model, relating these instrumental variables to the (unknown) true disease state is proposed. It is shown that this procedure allows the goodness-of-fit to the resulting model to be tested.