论文部分内容阅读
[目的]探讨具有连续型潜在变量的等级变量间多项相关系数、Pearson相关系数、秩相关系数间的关系。[方法]利用SAS统计软件产生样本数为10000的双变量正态分布,并将其转换成等级资料,分别求得其秩相关系数、Pearson相关系数及多项相关系数。[结果]存在正态分布潜在变量的等级资料的多项相关系数要比Pearson及秩相关系数更为接近实际的相关水平,特别当相关系数较大时,Pearson相关和秩相关系数与实际的偏差较大。[结论]多项相关系数受分级的影响不大,而Pearson相关和秩相关系数均受到分级的影响。
[Objective] To explore the relationship between multiple correlation coefficients, Pearson correlation coefficient and rank correlation coefficient among the rank variables with continuous latent variables. [Methods] The bivariate normal distribution with sample number of 10000 was generated by using SAS statistical software and converted into rank data to obtain rank correlation coefficient, Pearson correlation coefficient and multiple correlation coefficients respectively. [Results] The multiple correlation coefficients of the grade data with latent variables of normal distribution are closer to the actual correlation level than the Pearson and rank correlation coefficients, especially when the correlation coefficient is large, the Pearson correlation and rank correlation coefficient are different from the actual deviation Larger. [Conclusion] The correlation coefficients of multiple items are not affected by the grading. Pearson correlation and rank correlation coefficients are all affected by grading.