ESTIMATING THE RATE OF DEFECTS UNDER IMPERFECT SAMPLING INSPECTION—A NEW APPROACH

摘要

We consider the problem of estimating the incidence rate of events, givencounts of observations of these events made by two independent imperfectobservers during several time intervals. In contrast with the well-knownproblem of Multiple Systems Estimation (also known as Capture-Recapture), herewe do not have information regarding the number of events which are reported byboth observers. We solve this problem by constructing two types of estimators -a simple moment-type estimator, and a more complicated maximum-likelihoodestimator. The performance of these estimators is studied analytically and bymeans of simulations. It is shown that the maximum-likelihood estimator issuperior to the moment-type estimator. A systematic comparison with theMultiple Systems Estimation method is also made.