Simulation for Research
This book is about simulation modeling, programming and experimentation for the purpose of systems analysis. However, stochastic simulation is also a tool that can be used to support basic research in domains such as simulation, optimization, queueing, fi
- PDF / 149,696 Bytes
- 10 Pages / 439.36 x 666.15 pts Page_size
- 56 Downloads / 159 Views
Simulation for Research
This book is about simulation modeling, programming and experimentation for the purpose of systems analysis. However, stochastic simulation is also a tool that can be used to support basic research in domains such as simulation, optimization, queueing, financial engineering, production planning and logistics. In this chapter we cite some papers that demonstrate effective application of simulation in research and use them to highlight general principles and practices. We start with two important distinctions. The first distinction is between a practitioner’s experiment and a researcher’s experiment. This book is written primarily from the point of view of the practitioner’s experiment. The practitioner has a real problem to solve and has some limit on how much time or effort they can expend to solve it. The practitioner will try to build an appropriate simulation model or models, will perform the simulation experiment using the best methods they know or have access to, and will use the results to solve their problem. If they learned simulation from this (or many other) books then they will also try to assess the quality of their answer (e.g., via a confidence interval) or they will use a procedure that is designed to deliver a certain level of quality (e.g., a ranking and selection procedure with a guaranteed probability of correct selection). They will, however, only have a vague sense as to whether they got the “right answer” to their problem based on what happens when they apply their simulation results in the real world. The researcher’s experiment, on the other hand, is driven by an abstract research question whose answer may be useful in practice, but is not the answer to a specific practical problem. Because it is not directly tied to the real world, which is messy, a research question can be formulated and answered very precisely. The researcher’s experiment may, and often will, involve solving the same problem repeatedly for simulated instances that differ through some controllable experimental factors and in the random numbers used in the simulation. For instance, Sect. 9.1 below describes generating random optimization problems of a particular type and solving each instance with the same algorithm to assess the algorithm’s performance over a space of problems.
B.L. Nelson, Foundations and Methods of Stochastic Simulation: A First Course, International Series in Operations Research & Management Science 187, DOI 10.1007/978-1-4614-6160-9 9, © Springer Science+Business Media New York 2013
243
244
9 Simulation for Research
Unlike the practitioner’s experiment, the researcher’s experiment frequently involves simulating scenarios for which the answer is known in advance; knowing the answer facilitates a better evaluation. For example, Sect. 9.3 describes an experimental study of new ranking and selection procedures. Ranking and selection procedures are supposed to discover the best scenario with a guaranteed probability of correct selection. The achieved probability of correct sec
Data Loading...
