I have been working with the new RSABE template and the progesterone data set for a full replicate design (ABAB, BABA) provided by Ana Henry. I find it works well. Recent methylphenidate guidance indicates new requirements. One of them is the calculation of within TEST variance (WT) for use in other calculations.
As the RSABE is provided only the WR is calculated. It was suggested to me by Helmut Schutz that the way to get this information was to copy and paste “Prepare Data Sets for Analysis……….” Object in the RSABE workflow thereby creating a copy of “Prepare Data Sets for Analysis….” dedicated to providing (WT). I accepted this suggestion so I attach a copy of this Phoenix file.
Within the “Copy of Prepare data sets for analysis…..” there is an Object entitled Dij complete rows Data Wizard. This is the point where incomplete data is excluded by the transformation.
Since this work is focused on TEST not reference for the transformation I changed over to LOG DATA T1 for x column and LOG DATA T2 for the Y column
The exclusion criteria are below.
Exclude where [LOG DATA R1] is NULL entire ROW
Exclude where [LOG DATA R2] is NULL entire ROW
When I run the program I find it uses 69 subjects and my value for WT is 0.1186. This value differs from another workers SAS value (0.11654) using the same data set as posted to the Bioequivalence and Bioavailability Forum. Helmut himself has confirmed the SAS value in Phoenix. The question arises as to why my value is different. The SAS worker sent me his list of 71 subjects used by SAS for his WT determination. I note that two subjects (24 and (31) are not included in my data set. I am thinking that the above subject exclusion criteria may be incorrect.
I am thinking what do I have to do to get the same result for WT as the other workers.
OK I took a quick look at this (cross-reference for those who want the background from BEBAC forum http://forum.bebac.at/mix_entry.php?id=14150#14150). .%C2%A0)And as far as I can see these subjects 24 and 31 are dropped; correctly according to the filter applied in the step “Dataset ready for analysis_all_R_Periods”.
So at first glance I would say you have 73 Subjects who completed R1 & R2
(and 71 who completed T1 & T2, and 69 who completed all periods)
So I agree with you & Helmut on that and I am not sure why you are getting differing results overall to Helmut ?
maybe someone can repeat the question to me because I’m confused. What are you trying to calculate and why do you think it’s wrong in your modification of the template?
and if we look at Posting: # 14273 I can get the same number as Helmut for T for either 0.95 or 0.05, I’m not sure which one you want at this time?
Dependent Units Parameter Estimate INVCHI95 INVCH05
ilat Var(Residual) 0.16589777 87.108072 49.16227
the one you present above is the ratio parameter: I have that value and 49.16227 is the value I have too. { Helmut corrected that in a later post}. If you look at my "copy of prepare data sets for analysis…) the parameter I am after is dlat, which is the WT variance and it is 0.1186. Look at the object within this workflow in the “Swt and dfd” then you will the value for WT (within subject variance) is 0.1186. This is the parameter that is different from the other workers (0.116539674). I think it is due to the number of subjects included in the calculation. Here is my post to John on bebac,
Hi Angus, don’t forget you can edit an exisitng post if you just want to add an extra note.
I’m still confused; post Posting: # 14273 on BEBAC;
I would get 49.16227 for p 0.05 instead of 0.95. Maybe I got sumfink wrong? dfd Var_wt Cinv from workflow step… dlat(T) 69 0.116539674 89.39120787 > ??? dlat(R) 71 0.199313551 91.67023918 > Prepare …_s2wr and dfd ilat(T–R) 67 0.165897781 87.10807220 > Copy Prepare…_chiinv
so I too don’t know where John and Helmut got dlat(T) with a dfd of 69 and varWt 0.1165 from, surely it should be 71 for degrees of freedom since N=73 for T ?
Simon: the degrees of freedom the program is using for the calculation I make (to get 0.1186) is 67. That means I am using 69 subjects completed. John has done his own programming in SAS. His SAS work includes subject 24 and 31 and my work does not. He is including 2 more subjects that Phoenix excludes by the program. Regarding your 3 question marks at dlat Helmut’s number is incorrect for chi at that point: he later corrected it to m~49, which I agree with. here is my post to John today 30 th January 30, 2015.
I am thinking that the source of the problem may be that when preparing the “copy of data sets for all analysis” object that I did this in an alternate way from what should have been done. What I did fro the file I attached was I went to the RSABE object and double clicked on that to see the object “prepare data sets for analysis” at that point I copied it and that is the copy. It may have been intended that I copy the “Prepare Data sets for analysis…” without double clicking on RSABE… as you see it earlier. Then that copy must be changed extensively so that it accommodates the WT calculation instead of the WR so that you get the WT variance. What makes me say that is an earlier date Helmut warned me that all of the transformations, filters had to be changes. I wondered about this since the approach I used that you see did not need a lot of changes. But copying the “Prepare Data sets for analysis…” without double clicking on RSABE creates an object that requires a lot of changes. to the code. So I went to the "prepare data set for analysis’ object that comes standard with the RSABE and I modified it to “Exclude where [LOGDATAT1] IS NULL AND EXCLUDE where LOGDATAT2] is null” That way I got subject 24 and 31 included in the data set. I looked through the code in the other objects and I eliminated other exclusion criteria entries as I felt appropriate I changed the ratio over to including T1 and T2 rather than R1 and R2. Then when I implemented the analysis for WT (within subject variance) I got 0.1165394 and the Chi was 89.391268 (almost identical values to Helmut). Now I am thinking about the best way to structure the worksheet to provide sections for both WT and WR plus variance for the ratio so that you do not modify the WR part to get WT while you still get the variance for the ratio conveniently.
