关于空模型：组内相关系数以及模型的选择-经管之家官网！

各位前辈，在下想咨询一个关于多水平模型的问题，我的目的是想建立一个社区-个体-调查时间的三水平空模型，看看建立三水平模型分析是否合适。在stata上写语句 . xtmelogit hamin2||COMMID:||IDind:
得到的结果是：

1. Refining starting values:
   
2. 
   
3. Iteration 0: log likelihood = -2930.1485
   
4. Iteration 1: log likelihood = -2894.2116
   
5. Iteration 2: log likelihood = -2868.7771
   
6. 
   
7. Performing gradient-based optimization:
   
8. 
   
9. Iteration 0: log likelihood = -2868.7771
   
10. Iteration 1: log likelihood = -2868.6202
    
11. Iteration 2: log likelihood = -2868.6199
    
12. Iteration 3: log likelihood = -2868.6199
    
13. 
    
14. Mixed-effects logistic regression Number of obs = 5430
    
15. 
    
16. --------------------------------------------------------------------------
    
17. | No. of Observations per Group Integration
    
18. Group Variable | Groups Minimum Average Maximum Points
    
19. ----------------+---------------------------------------------------------
    
20. COMMID | 308 1 17.6 66 7
    
21. IDind | 3998 1 1.4 3 7
    
22. --------------------------------------------------------------------------
    
23. 
    
24. Wald chi2(0) = .
    
25. Log likelihood = -2868.6199 Prob > chi2 = .
    
26. 
    
27. ------------------------------------------------------------------------------
    
28. hamin2 | Coef. Std. Err. z P>|z| [95% Conf. Interval]
    
29. -------------+----------------------------------------------------------------
    
30. _cons |-2.272335 .1503214 -15.12 0.000 -2.566959 -1.97771
    
31. ------------------------------------------------------------------------------
    
32. 
    
33. ------------------------------------------------------------------------------
    
34. Random-effects Parameters| Estimate Std. Err. [95% Conf. Interval]
    
35. -----------------------------+------------------------------------------------
    
36. COMMID: Identity |
    
37. sd(_cons) | 1.31052 .1211253 1.09338 1.570783
    
38. -----------------------------+------------------------------------------------
    
39. IDind: Identity |
    
40. sd(_cons) | 2.002017 .1565074 1.717612 2.333513
    
41. ------------------------------------------------------------------------------
    
42. LR test vs. logistic regression: chi2(2) = 448.62 Prob > chi2 = 0.0000
    
43. 
    
44. Note: LR test is conservative and provided only for reference.
    
45. 
    
46. .

复制代码
我在书上看这种分三层的模型在随机效应结果处会出现三行sd，然后用每一层的σ除以三个相加的sd就能得到ICC，但为什么我的结果里面只有二三水平的sd？是默认第一水平sd为1？不知怎么算ICC，如果不算icc就无法知道三水平是否合适呀。
另外我也想请教关于两水平和三水平模型的比较，做lrtest这样可以吗？显然三水平模型的likelihood增加了，这样怎么说明三水平模型优于二水平呢？