Inter-subject CV calculation for a Crossover Design

感谢神：新年伊始便让我有论坛币来象大家请教！
It is well-known that the inter-subject cv of a crossover design can be derived from the variance estimate of the SUBJECT effect. For example,

1. data test;
   
2. input Subject Period Treatment $ Sequence $ AUCt;
   
3. lnAUCt = log(AUCt);
   
4. datalines;
   
5. 1 1 R RT 8551.8
   
6. 1 2 T RT 4538.3
   
7. 3 1 R RT 3498.7
   
8. 3 2 T RT 10710.5
   
9. 6 1 R RT 6744.1
   
10. 6 2 T RT 5244.8
    
11. 10 1 R RT 4723.4
    
12. 10 2 T RT 2275.2
    
13. 12 1 R RT 9048.8
    
14. 12 2 T RT 6669.6
    
15. 14 1 R RT 3817.9
    
16. 14 2 T RT 3841.7
    
17. 15 1 R RT 9801.1
    
18. 15 2 T RT 3286.6
    
19. 18 1 R RT 13406.3
    
20. 18 2 T RT 7070.3
    
21. 22 1 R RT 7644.5
    
22. 22 2 T RT 10794.6
    
23. 23 1 R RT 6152.1
    
24. 23 2 T RT 3448.3
    
25. 25 1 R RT 10842.8
    
26. 25 2 T RT 11058.5
    
27. 2 1 T TR 2199.6
    
28. 2 2 R TR 11036.4
    
29. 4 1 T TR 3419.9
    
30. 4 2 R TR 6447.8
    
31. 5 1 T TR 4523.4
    
32. 5 2 R TR 5190.9
    
33. 8 1 T TR 7043.7
    
34. 8 2 R TR 3786.6
    
35. 9 1 T TR 3948.0
    
36. 9 2 R TR 7159.8
    
37. 11 1 T TR 1768.7
    
38. 11 2 R TR 2380.9
    
39. 13 1 T TR 8922.0
    
40. 13 2 R TR 2443.8
    
41. 16 1 T TR 4159.6
    
42. 16 2 R TR 3217.2
    
43. 17 1 T TR 5566.6
    
44. 17 2 R TR 9262.4
    
45. 20 1 T TR 5030.5
    
46. 20 2 R TR 6086.9
    
47. 21 1 T TR 3989.6
    
48. 21 2 R TR 3019.1
    
49. 24 1 T TR 5519.4
    
50. 24 2 R TR 5583.8
    
51. 26 1 T TR 5879.7
    
52. 26 2 R TR 5539.5
    
53. ;
    
54. proc mixed data=test method=type3;
    
55. class subject sequence period treatment;
    
56. model lnAUCt = Sequence Period Treatment;
    
57. random Subject(Sequence);
    
58. run;

复制代码

The estimated intersubject-cv = sqrt(exp(0.009933)-1), where 0.009933 is obtained from the ODS output table 'Covariance Parameter Estimate'.
Note that the estimate is consistent regardless of the METHOD= option in the PROC MIXED statement for a balanced dataset.
However, different METHOD= options lead to different answers when the dataset is NOT balanced. For example, if you delete the last observation from the sample dataset, then methods TYPE1/TYPE3/ML/REML/MIVQUE will have different variance estimates.
Question: Is there a preferred METHOD= option in PROC MIXED for intersubject CV estimation? Why?

楼主，请问intra-subject variation怎么计算呢？尤其是higher-order crossover clinical trial,因为有预试验中，我想知道个体内变异估算样本量，谢谢！

ch03en12tong 发表于 2014-1-9 10:11
楼主，请问intra-subject variation怎么计算呢？尤其是higher-order crossover clinical trial,因为有预试验 ...

http://hansheng.gsm.pku.edu.cn/pdf/2002/power.pdf Equation (2)

邓贵大 发表于 2014-1-9 21:46
http://hansheng.gsm.pku.edu.cn/pdf/2002/power.pdf Equation (2)

谢谢楼主的回复，我好好学习下，如果有不懂的再向您请教！谢谢！