OLS与PLS方法在顾客满意度建模中的模拟比较

使用蒙特卡罗方法 ,对比研究了使用普通最小二乘法和偏最小二乘法 ,进行顾客满意度多元线性建模的效果。模拟结果表明 ,采用偏最小二乘法在样本量不足 ,设计阵存在多重共线性时能够获得更好的建模可靠性 ,而样本数据的右偏斜分布对这两种方法均不产生显著的危害

Comparison between OLS and PLS in Customer Satisfaction Modeling

Using Monte Carlo simulation, a comparison between OLS (Ordinary Least Squares) and PLS (Partial Least Squares) is performed in customer satisfaction modeling. The result shows that under the conditions of small sample and multi-collinearity, PLS can reach higher reliability than OLS do. However, under the condition of right skew distribution, both PLS and OLS will perform well.