The use of auxiliary functions in nonparametric inference for the M/G/∞ queueing model is considered. Estimation of the service time distribution G is challenging when only limited information about the busy/idle cycle is available. It is shown, using diagnostic plots of estimators, that a standard auxiliary function aimed at providing numerical stability fails in that regard but that a reliable auxiliary function can be constructed. The improvement made by the alternative auxiliary function is demonstrated.
The final, definitive version of this article has been published in the Journal, Computational Statistics and Data Analysis 51 (12), 2007, © ELSEVIER.