Vector computers, Monte Carlo simulation and regression analysis: an introduction

Article Abstract:

Vector computers are offering increasing supercomputer applications to management scientists. These applications are examined in terms of Monte Carlo experiments using regression analysis, taking into consideration the fact that these applications can only be facilitated through vector mode thinking. Three important observations are made regarding supercomputers. One observation relates to the necessity of using adjusted algorithms to increase the efficiency of supercomputing. It is also noted that for specific types of problems, supercomputers display slower functionality than the multipurpose machines. The last observation concerns the enhanced calculating speed offered by supercomputers. Although these supercomputers have the ability of facilitate calculations at greater speeds, the time spent on the construction of required codes may not compensate for this advantage.

Evaluation, Multivariate analysis, Supercomputers

---

Validation of trace-driven simulation models: a novel regression test

Article Abstract:

A novel regression test to validate trace-driven simulation models was proposed. The procedure allows the responses of the simulated and real systems to have the same means and variances, an alternative requirement to the concept that regressing the outputs of a trace-driven simulation approach on the observed real outcomes should produce a 45-degree line through the origin. The differences between simulated and real responses were regressed on their associated sums and the resulting slope and intercept were tested to determine statistically whether the alternative requirement is met. Results showed that the probability of rejecting a correct simulation model is significantly higher with the naive test than with the novel regression test. It was argued that it is wrong to regress simulated outputs of a trace-driven simulation on real outcomes.

Research and Development in the Physical, Engineering, and Life Sciences, Statistics, Models, Simulation methods, Simulation, Regression analysis, Statistics (Mathematics), Correlation (Statistics), Goodness-of-fit tests

---

Analyzing simulation experiments with common random numbers

Article Abstract:

A linear regression model with a non-diagonal co-variance matrix is used to analyze simulation runs which use common random numbers. It is not necessary for the co-variance matrix to have a specific pattern. A new framework for the error analysis is proposed which consists of three factors: common random numbers, replication, and model validity.

Management science, Error analysis (Mathematics), Linear models (Statistics)

Similar abstracts:

• Abstracts: Board composition and the commission of illegal acts: an investigation of Fortune 500 companies. Directors' characteristics and committee membership: an investigation of type, occupation, tenure, and gender
• Abstracts: Competition and cooperation: new trends in corporate public issue identification and resolution. Issues management: Corporate fad or corporate function?
• Abstracts: Cultural change: an integration of three different views. Adoption, discontinuation and retention of a capital good innovation
• Abstracts: Strategy and organization. Strategy and organisation
• Abstracts: Strategic considerations for unaudited account values in analytical review

This website is not affiliated with document authors or copyright owners. This page is provided for informational purposes only. Unintentional errors are possible.