zhens11 发表于 2010-1-14 10:48 
我运行mixed模型，运行结果感觉都还不错，不过在LOG里面出现以下语句：
Convergence criteria met but final hessian is not positive definite，
这是什么意思啊？
哪个地方出现问题了吗？
谢谢大家！
In theory, the hessian should be positive in this problem. Think about the inverse hessian is the estimated covarance matrix.
The message means
1) your model overspecified
2) fitting duplicate parameters when your specified effect/variance structure statements.
3) When you have random effects, the one of varance componements may be zero(almost zero) from estimation.
Basicly you could ignore the message if you understand what your are doing.
扫码加好友，拉您进群
全部版块
我的主页
收藏
