An Old Problem Revisited—and Solved: Upton and Dalton's Linear Production Response: A Note
提出一种稳健方法，用于估计由线性反应区间和线性平台组成的反应曲面，该方法基于贝叶斯切换回归模型，并展示了吉布斯采样和数据增强的常规应用。
This note presents a robust method for estimating response surfaces that consist of linear response regimes and a linear plateau. The linear response‐and‐plateau model has fascinated production scientists since von Liebig (1855) and, as Upton and Dalton indicated, some years ago in this Journal the response‐and‐plateau model seems to fit the data in many empirical studies. The estimation algorithm evolves from Bayesian implementation of a switching‐regression (finite mixtures) model and demonstrates routine application of Gibbs sampling and data augmentation—techniques that are now in widespread application in other disciplines.