Using monte carlo simulation to choose among alternative statistical definitions
Winter Simulation Conference(2005)
摘要
We frequently need to come to agreement with our customers about the formulas that will be used to calculate various statistics relating to the measurement of contractual terms and conditions. In many cases there is only one generally accepted operational definition, but in other cases, such as computing percentiles, there are multiple definitions that can be considered. This presentation looks at work I have done using Monte Carlo simulation to compare two different formulas for estimating a population percentile from small samples; the application is the analysis of benchmarking data. This work compares two different percentile calculations by comparing the performance of both approaches under multiple probability models with respect to estimation of the 10th, 25th, 50th, 75th, and 90th percentiles. In addition, a non-parametric confidence interval generation method for each of these percentiles was also examined to determine how well it performed relative to its stated 95% confidence level.
更多查看译文
关键词
monte carlo simulation,confidence level,different percentile calculation,benchmarking data,population percentile,different formula,non-parametric confidence interval generation,contractual term,multiple probability model,multiple definition,alternative statistical definition
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络