3 ESTIMATORS OF THE PROBABILITY OF OPIATE USE FROM INCOMPLETE DATA

Three estimators of the probability of opiate use are developed involving correlated binomial random variables. These estimators differ as different patterns of correlation are considered with regard to the use of the opiate on successive visits by the same study subject. These estimators also allow for subjects to miss clinic visits wherein the purpose of these visits is to test the subject for the presence of the opiate. The procedures developed are then applied to the data arising from the double-blind three-armed controlled ARC 090 trial.