Exploring exponentially weighted moving average control charts to determine the warm-up period

摘要

In this paper, we examine the use of exponentially weighted moving average (EWMA) control charts for the detection of initialization bias in steady state simulation experiments. EWMA charts have the interesting property of being more sensitive to shifts in the data as compared to other control charting techniques. We exploit this sensitivity by developing a criteria for searching for the deletion point when the EWMA is applied to the reversed data sequence. This allows us to more easily detect and count the number of times the smoothed sequence remains in control. Our results indicate that the procedure can quickly find and recommend a deletion point. In addition, the properties of the resulting estimators are good if the dataset that is being analyzed does not have an overtly large amount of biased data points. We use experimental test cases to illustrate the properties of the procedure.