Consider the problem of searching for a leprechaun that moves randomly between two sites. The movement is modelled with a two-state Markov chain. One of the sites is searched at each time t = 1,2,…, until the leprechaun is found. Associated with each search of site i is an overlook probability αi and a cost Ci Our aim is to determine the policy that will find the leprechaun with the minimal average cost. Let p denote the probability that the leprechaun is at site 1. Ross conjectured that an optimal policy can be defined in terms of a threshold probability P* such that site 1 is searched if and only if p ≥ P*. We show this conjecture to be correct (i) when α1 = α2 and C1 = C2, (ii) for general Ci when the overlook probabilities α, are small, and (iii) for general αi and Ci for a large range of transition laws for the movement. We also derive some properties of the optimal policy for the problem on n sites in the no-overlook case and for the case where each site has the same αi, and Ci.

In this paper we present a unified approach to the optimal dimensioning and tariffing of loss networks. In our formulation the optimum is chosen to maximize the profit for the company operating the loss network. We assume that the operating company has the flexibility to determine tariffs and grade of service — although both of these can possibly be subject to regulatory constraints. The fact that the tariffing may affect demand and, hence, the dimensioning makes it essential that the operating company include the tariff/demand trade-off in determining the optimal way to dimension the loss network. A consequence of our formulation is that the optimal tariff structure has a particularly simple form, with the optimal tariff on a particular route separating into a term related to the tariff/demand trade-off on that route and a term that reflects the cost of the circuits used by the route.

Gerald J. Lieberman was born on December 31, 1925, in Brooklyn, New York, after a hectic New Year's Eve trip to the hospital. His father, Joseph, spelled his last name Liberman, but his mother, Ida, preferred Lieberman, the spelling that she and some of Joseph's siblings used. Joseph and Ida had come to this country from Lithuania. Joseph worked for the Metropolitan Life Insurance Company, and they lived in an “historic” section of Flatbush. The much wanted baby boy was the center of the family, which included two doting older sisters, Shirley and Rosalind. He grew fast — one of the tallest boys in nearby Public School 197 — and achieved his adult height at about the age of 13. As a boy, he was described as towheaded and gawky. Jerry did not realize that he had a middle initial until he was 15 and needed a birth certificate to get a work permit. Jerry asked his parents if they had given him a middle initial, but they did not remember. In any case, since the J does not stand for anything, Jerry likes to quip that his middle name is Jinitial.

This paper describes a generalization of the classical regenerative method of simulation output analysis. Instead of blocking a generated sample path on returns to a fixed return state, a more general scheme to randomly decompose the path is used. In some cases, this decomposition scheme results in regeneration times that are a supersequence of the classical regeneration times. This “accelerated” regeneration is advantageous in several simulation contexts. It is shown that when this decomposition scheme accelerates regeneration relative to the classical regenerative method, it also yields a smaller asymptotic variance of the regenerative variance estimator than the classical method. Several other contexts in which increased regeneration frequency is beneficial are also discussed.

Via order statistics we analyze the average length of all separators in random Prefix B-trees. From this result we draw some conclusions and conjectures concerning the average overall storage of random Prefix B-trees.

A simulation method is developed for computing average reward optimal policies, for a finite state and action Markovian decision process. It is shown that the method is consistent; i.e., it produces solutions arbitrarily close to the optimal. Various types of estimation errors and confidence bounds are examined. Finally, it is shown that the probability distribution of the number of simulation cycles required to compute an e-optimal policy satisfies a large deviations property.

This paper extends product form results for queueing networks with signals to allow history-dependent routing. The signals in these models carry information to nodes and induce multiple customers to move simultaneously within the network. Two models are studied in this paper. In the first one we assume that routing probabilities of a departing customer from a given class of nodes depend on the amount of service just received by the customer and whether its departure is the result of an actual service completion or the result of an arriving signal. In the second model we assume that the routing probabilities of a customer depend on the number of times this customer's service has been interrupted by signals in the past as well as the cause of its departure. We show that both models possess simple product form solutions. These results provide a new dimension in modeling and analyzing practical systems.

In the fall of 1948 in my course on Least Squares in the Department of Mathematical Statistics at Columbia University (and in the spring in Correlation and Chi-Square), I was particularly impressed by one of the students— Gerald J. Lieberman. I was disappointed that this promising student left Columbia after one year, but it was not long until our paths met again. It is a pleasure to dedicate this paper to my colleague and close friend!

Consider an M/G/1 type queueing system with two service modes: regular speed and high speed. The service rule is characterized by two switch-over levels nR and nH, where nR and nH are given integers with 0 ≤ nH < nR. The server switches from regular speed mode to high speed mode when the number of customers present at a service completion epoch is equal to or larger than nR and switches from high speed mode to regular speed mode when the number of customers present decreases to nH. A key feature of the model is that the server takes a vacation for setup operations before a new service mode is available. This paper derives for the general model an expression for the generating function of the equilibrium queue-length distribution in terms of the switch-over levels. Two unknown parameters appear in the generating functions. Using a recursive method, we solve these unknown parameters and obtain a computationally tractable algorithm for the steady-state probabilities.

In this note we consider a single-server queueing loss system with zero buffer. The arrival process is a nonstationary Markov-modulated Poisson process. The arrival process in state i is Poisson with rate λi. The process remains in state i for a time that is exponentially distributed with rate Cαi, with c being a control or speed parameter. The service rate in state i is exponentially distributed with rate μi. The process moves from state i to state j with transition probability qij. We are interested in the loss probability as a function of c. In this note we show that, under certain conditions, the loss probability decreases when the c increases. As such, this result generalizes a result obtained earlier by Fond and Ross.

The polling network considered here consists of a finite collection of stations visited successively by a single server according to a state-dependent routing scheme. At every visit of a station, a possibly random number of customers are served and depart from the network thereafter. The server takes a possibly random time to walk from one station to the next. The network receives groups of customers at Poisson instants. These customers are distributed randomly over the stations and wait until served. A necessary condition for stability is derived, which proves to be also sufficient in certain special cases. Such cases include a starlike network and a ringlike one with a greedy service policy. Mathematically, positive recurrence of a Markov chain is studied.

We present a new exponential inequality for degenerate U-statistics. The bound on the tail probability is quadratic for small to medium values of the deviation and linear for larger values. We apply this bound to a family of test statistics and provide the key step in an optimality result for adaptive tests [1].

We study the problem of broadcasting in a system where nodes are equipped with radio transmitters with constant radius of transmission. A message originating at a node has to be transmitted to all the other nodes in the system. We prove the central limit theorem and the law of large numbers for the number of time steps required to complete a broadcast for the case when the nodes are placed on a line independently uniformly distributed. We show that the number of time steps required to broadcast is 3n/4 in probability.

Let X be a random variable with characteristic function ϕ. In the case where X is integer-valued and n is a positive integer, a formula (in terms of ϕ) for the probability that n divides X is presented. The derivation of this formula is quite simple and uses only the basic properties of expectation and complex numbers. The formula easily generalizes to one for the distribution of X mod n. Computational simplifications and the relation to the inversion formula are also discussed; the latter topic includes a new inversion formula when the range of X is finite.

When X may take on a more general distribution, limiting considerations of the previous formulas suggest others for the distribution, density, and moments of the fractional part X — [X]. These are easily derived using basic properties of Fourier series. These formulas also yield an alternative inversion formula for ϕ when the range of X is bounded.

Applications to renewal theory and random walks are suggested. A by-product of the approach is a probabilistic method for the evaluation of infinite series.

Let x = (x1, x2, …) be an infinite sequence of 0's and 1's, initially unknown to you. On day n = 1,2,…, you observe hn = (x1, …, xn–1), the first n – 1 terms of the sequence, and must predict xn. What is a good prediction method, and how well can you do?

Motivated by extreme-value engineering in service systems, we develop and evaluate simple approximations for the distributions of maximum values of queueing processes over large time intervals. We provide approximations for several different processes, such as the waiting times of successive customers, the remaining workload at an arbitrary time, and the queue length at an arbitrary time, in a variety of models. All our approximations are based on extreme-value limit theorems. Our first approach is to approximate the queueing process by one-dimensional reflected Brownian motion (RBM). We then apply the extremevalue limit for RBM, which we derive here. Our second approach starts from exponential asymptotics for the tail of the steady-state distribution. We obtain an approximation by relating the given process to an associated sequence of i.i.d. random variables with the same asymptotic exponential tail. We use estimates of the asymptotic variance of the queueing process to determine an approximate number of variables in this associated i.i.d. sequence. Our third approach is to simplify GI/G/1 extreme-value limiting formulas in Iglehart [25] by approximating the distribution of an idle period by the stationary-excess distribution of an interarrival time. We use simulation to evaluate the quality of these approximations for the maximum workload. From the simulations we obtain a rough estimate of the time when the extreme-value limit theorems begin to yield good approximations.

A general type of model that results in components having a negative type of dependence is proposed. Specifically, a system of dependent binary components, in which component i is said to function if at least one of a given number of multinomial trials results in that outcome, is considered. The multinomial probabilities that maximize the probability of a k-of-n system are determined. Some approximations for the system reliability are proposed first for a system that functions unless k consecutive components are failed and then for a general system. A simulation method that uses a variance reduction technique to estimate the reliability is presented.

We present a new estimator for the standard deviation parameter in the standard quality control model. We argue that it has a smaller mean square error than the usual estimator when the data are normal and that, unlike the usual estimator, it remains consistent even when the data are not normal.

Let N = [N(t), t ≥ 0] be a point process, and let T1 and T2 be two positive random variables that are independent of N. Define Mk = N(Tk), k = 1,2. In this paper we study conditions on the process N under which some stochastic orders between T1 and T2 are transformed into stochastic orders between M1 and M2. In some cases the converses are also shown to be true. Some applications and examples are given.

We give a computable bound on the rate of convergence of the occupation measure for the Gibbs sampler to the stationary distribution.