Mitigating the Effects of Multicollinearity Using Exact and Stochastic Restrictions: The Case of an Aggregate Agricultural Production Function in Thailand
比较了普通最小二乘法、精确约束OLS、随机约束OLS（混合估计）和主成分回归在估计泰国农业总生产函数时的表现，发现后两种方法能有效缓解多重共线性问题，并优于OLS。
Abstract Ordinary least squares, exactly restricted OLS, stochastically restricted OLS (mixed estimation), and principal components regression each were used to estimate an aggregate agricultural production function for Thailand for which data were highly multicollinear. Pretest considerations, incorporating alternative risk measures, were addressed in detail for purposes of model evaluation. The final mixed and principal components models generally outperformed OLS in terms of risk and overall reasonableness, mitigating a serious multicollinearity problem and permitting a direct examination of the rate and composition of Thai agricultural output growth.