| Summary: | 碩士 === 國立高雄大學 === 統計學研究所 === 97 === Response surface methodology(RSM) is a popular method of searching optimal setting. Typically, researchers apply central composite design(CCD) with repeated measurements at center point to estimate the variance of the response. Then consider the analysis of variance(ANOVA) by assume each of the response is generate from a normal distribution with constant variance. However, if we have no repeat measurement under a same setting, ANOVA can not be done. In this case, Chen et al. (2006) gave a test statistic SC which can be used to detect whether the curvature is significance in the model. In this thesis, we follow the idea of them to create another test statistic CS and give its distribution properties under certain assumptions.
|
|---|
