Abstract:
|
In the last twenty years, thousands of response surface studies have been published. Using the Web of Science Application Program Interface, we extracted over 20,000 records of journal articles from all Science Citation Index disciplines, based upon keywords indicating that they involved a response surface study. We secured a random sample of more than 100, stratified by the number of factors, and obtained the design as well as information about factors and other characteristics of the experiments. These designs were then reanalyzed and the results gathered together in a meta-analysis to reveal information about effect heredity, hierarchy, and sparsity. We empirically quantify these principles to guide researchers toward more realistic simulation scenarios and more efficient designs, and provide evidence for what practitioners have informally observed.
|