急，交叉分析，help me

subject perieod sequence treatment AUC
1       1 AB A 3981.76
2       1 AB A 5573.62
3       1 AB A 5821.52
4       1 AB A 8558.99
5       1 BA B 3958.93
6       1 BA B 5145.91
7       1 BA B 689.6
8       1 BA B 3812.02
1       2 AB B 4446.05
2       2 AB B 4847.12
3       2 AB B 4274.72
4       2 AB B 4839.49
5       2 BA A 3219.92
6       2 BA A 4587.32
7       2 BA A 4436.7
8       2 BA A 6168.8
像上面的crossover有两组，如果subject(sequence) 为random effect,那么比较两个treatment可以这样：
proc mixed data=test;
  class treatment period sequence subject;
  model AUC=treatment period sequence
  random subject(sequence);
Run;
当然用proc glm也一样可以搞定，但是现在有2个人是对照组，即
9  1 C C 4302.5
10 1 C C 5321.1
9  2 C C 4731.5
10 2 C C 5482.9
现在我要分析比较3个组，比较A,B,C 3组，求出sequence，treatment，period的P值，这个应该当成什么来做？或者data结构要变吗？repeat anova? or crossover anova?

It becomes no crossover dessign no more by adding subject 9/10. Many terms are non-estimable. For example, you cannot distinguish the Period effect and the carry-over effect for Treatment C.

邓贵大 发表于 2014-3-26 05:06
It becomes no crossover dessign no more by adding subject 9/10. Many terms are non-estimable. For ex ...

The option HTYPE= after the MODEL statement has valid values 1,2,3, could you tell me the difference of 1,2,3?