Abstract
A simulation study comparing a set of alternatives should include a confidence statement regarding the accuracy of its results. Statistical ranking and selection procedures can provide such confidence statements. Typical goals for these procedures in clude selecting the best alternative, selecting a subset that includes the best, selecting all alter natives that are better than a control (or base case), selecting all alternatives that are better than a control while also excluding all alternatives that fail to meet a minimum standard, and selecting all alternatives that are better than a standard involving two parameters.
The analyst can use tables to determine how many observations are required to achieve a specified level of confidence in the results. Proportions are generally the best statistics to use because they do not assume any underlying distribution for the data and, in some cases, they summarize distributions more effectively than do such other parameters as the mean or median. Examples show how the procedures were applied in an evaluation of remote-access inter active computer services.
Keywords
Get full access to this article
View all access options for this article.