Peter Glynn, Stanford University
In this talk, we will discuss how statistical modeling and insights gained from limit theorems for queues can be fruitfully used in combination with one another to improve decision-making, as illustrated by many server queueing applications. In particular, the rich stochastic modeling literature identifies the key statistical features in the underlying observed data that drive performance in such systems, and this impacts the types of statistical models that one should adopt. In addition, considerations related to computational, analytical, and statistical tractability shape one’s modeling choices. This talk will discuss this modeling perspective, and some of the recent theoretical, modeling, and computational tools that support this framework.