Log in
Register|
jag009 ★★★ NJ, 2013-08-28 22:13 (4276 d 19:45 ago) Posting: # 11374 Views: 6,414 |
|
|
Hi all, Is there a way to extract the intersubject CV from Proc Mixed for a 3-way partial replicate study (Based on the ABE part of the FDA RSABE Proc Mixed code)? Covariance Parameter EstimatesThanks John |
|
d_labes ★★★ Berlin, Germany, 2013-08-29 16:54 (4276 d 01:03 ago) @ jag009 Posting: # 11384 Views: 5,333 |
|
|
Dear John, ❝ Is there a way to extract the intersubject CV from Proc Mixed for a 3-way partial replicate study (Based on the ABE part of the FDA RSABE Proc Mixed code)? From the FA0(2) parametrization there are complicated formulas conecting the covariance parameters with the intersubject CV's. A simpler way would be to request the output of the G-Matrix via RANDOM treat/TYPE=FA0(2) SUB=subject G;If you used the FDA code the G-Matrix is already requested This matrix can be interpreted as ( s2bT rho*sbT*sbR ) where the diagonal elements (inter-subject variances) can be converted to the CV's via CVbT=sqrt(exp(s2bT)-1)Cave on the order in the G-Matrix. It depends on your coding for treatment. BTW: This way is not restricted to the partial replicate design but is general. But in case of it the trust in the numbers is restricted (at least by me  ).— Regards, Detlew |
|
jag009 ★★★ NJ, 2013-08-29 17:09 (4276 d 00:49 ago) @ d_labes Posting: # 11385 Views: 5,242 |
|
|
Thank Detlew, The output I first posted is from the FDA codes and yes "RANDOM treat/TYPE=FA0(2) SUB=subject G" is already there. ❝ This matrix can be interpreted as ❝ ❝ ❝ where the diagonal elements (inter-subject variances) can be converted to the CV's via ❝ ❝ ❝ ❝ Cave on the order in the G-Matrix. It depends on your coding for treatment. Use the example output I posted originally, can you walk me through the above maze? On another note (off topic but it's short so I might as well...) if you have a 2x2x2 crossover study carried out in 2 groups, would the following Proc GLM model statement be sufficient? The only one in the model statment which I have question on is PERIOD(GROUP). MODEL &pk= GROUP SEQUENCE SUBJECT(GROUP*SEQUENCE) PERIOD(GROUP) TREATMENT GROUP*TREATMENT/SS1 SS3;Thanks John |
|
d_labes ★★★ Berlin, Germany, 2013-08-30 11:56 (4275 d 06:02 ago) @ jag009 Posting: # 11394 Views: 5,170 |
|
|
Dear John, ❝ Use the example output I posted originally, can you walk me through the above maze? I don't see the G-matrix in your post. Need I glasses ?Have also a look at this post. ❝ On another note (off topic but it's short so I might as well...) if you have a 2x2x2 crossover study carried out in 2 groups, would the following Proc GLM model statement be sufficient? The only one in the model statment which I have question on is PERIOD(GROUP). ❝ ❝ ❝ I would agree. But: Barbara Davit (in a letter to one of my clients) additionally to your code requested the group-by-sequence interaction in the model. Edit: See another thread – starting with this post. Helmut — Regards, Detlew |
|
jag009 ★★★ NJ, 2013-09-04 19:29 (4269 d 22:29 ago) @ d_labes Posting: # 11438 Views: 5,191 |
|
|
Hi Detlew, ❝ I don't see the G-matrix in your post. Need I glasses ❝ Have also a look at this post. Thanks! Helmut probably would chase me around with a broadsword for not searching the forum for previous postings.  ❝ I would agree. But: ❝ Barbara Davit (in a letter to one of my clients) additionally to your code requested the group-by-sequence interaction in the model. I guess she has different opinion. The code I listed was requested by FDA a while back for one of my old studies (actually my boss' belonging, 15 yrs ago). John |
Ing. Helmut Schütz