| Nelson and Staum derived ranking-and-selection procedures that employ control-variate (CV) estimators instead of sample means to obtain greater statistical efficiency. However, control-variate estimators require more computational effort than sample means, and effective controls must be identified. In this dissertation, we present a new CV screening procedure to avoid much of the computation cost along with a better paired CV model than Nelson and Staum. We also present a two-stage CV combined procedure which captures the ability to eliminate inferior systems in the first stage and the statistical efficiency of control variates for selection in the second stage. Some guidelines about control-variate selection and an empirical evaluation are provided. Kim and Nelson (2001) proposed a fully sequential procedure that can be more efficient than those two-stage procedures. We develop the fully sequential procedures that can take advantage of control variates. Some empirical evaluation and an illustration are also provided. |
