SEM基于相关系数矩阵,你只要定义好其就可以用了

ls的朋友可否解释得再详细些，跪谢了！

其实分类数据可以变换，通过变换就可以直接计算相关系数

I would try mixture model incorporating distributions for categorical data: negative binomial, poisson, etc. This can be considered as latent class analysis - each cluster or latent class would have different distribution.

谢谢楼上的朋友，这个问题已经解决了。