您这问题问的……太笼统啦。Mixed模型只要求Y （因变量）是正态分布。
很多模型Proc Mixed都可以做。

1) 您这问题问的……太笼统啦
Agree

2)Mixed模型只要求Y （因变量）是正态分布。
Wrong.  Which textbook tells you the response needs to be normally distributed?
The classical regression only needs the assumptions of moment conditions for error term, usually the first and second moments.
and error term is not correlated with any covariates.
E(err)=0;
E(err^2)=sigma^2;
E(err * x) =0;

3)很多模型Proc Mixed都可以做
True.

The mixed procedure deals with problems with repeated measures or panel data. With this data structure one can assume more flexible or more realistic model. It provides much more variance -covariance structures one could estimate and test.
a) it fits a random effect model/a random coefficient model ( coefficients are random draw distributions. Proc mixed assumes the distribution is normal currently--limitation)
b) it fits a fixed effect model with repeated measures -- which is much like proc glm

请问proc mixed可以求描述性统计量吗