msmnainar
★    

India,
2017-01-20 08:49
(2625 d 00:40 ago)

Posting: # 16985
Views: 5,797
 

 LSM differences in SAS vs Phoenix [Software]

Dear all

Could you please clarify the reason for getting results of LSM differences and confidence limit values in opposite to one another (eg., negative vs positive or positive vs negative values) when you evaluate the same data using SAS and Phoenix WinNonlin.


Thanks

Sundar. M
Shuanghe
★★  

Spain,
2017-01-20 13:31
(2624 d 19:58 ago)

@ msmnainar
Posting: # 16990
Views: 4,972
 

 LSM differences in SAS vs Phoenix

Hi Sundar,

From what you said, I guess you should check your SAS code first. Maybe you coded treatment as A and B for test and reference but you copy/paste the sas code from guideline or somewhere else? If so, change the order of -1 1 to 1 -1 in estimate statement since many of those code assume that the treatment is coded as T and R, so the order is reversed. e.g.ESTIMATE 'Test vs. Reference' treat -1 1 to ESTIMATE 'Test vs. Reference' treat 1 -1.

If your sas code is correct then I couldn't think of any reason how one can screw up Phoenix setting since you would visually assigned which formulation is reference, whether it is coded as A/B, T/R or 1/2. You should know which is which.

All the best,
Shuanghe
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2017-01-20 14:34
(2624 d 18:55 ago)

@ msmnainar
Posting: # 16991
Views: 5,043
 

 Lexical order

Hi Sundar,

❝ Could you please clarify the reason for getting results of LSM differences and confidence limit values in opposite to one another (eg., negative vs positive or positive vs negative values) when you evaluate the same data using SAS and Phoenix WinNonlin.


Both software packages by default give the LSM difference in lexical order. If you coded test with T and reference with R, the difference will be given as R – T (since R < T).
It might be that your SAS-code changed the order by a statement similar to this one:

lsmeans formulation / pdiff=control("R") CL alpha=0.10;

In Phoenix there is no native method the change the order which you will get in
Bioequivalence > Output Data > LSM Differences.

[image]


My method to recode the LSM differences (which works for any specification of T and R):
LSM Differences > Send To > Data > Data Wizard
Rename to Reordered LSM Differences and add these Custom Transformations:
  1. R: left(Level_Level, search(Level_Level, '-', 1)-2)
  2. T: right(Level_Level, search(Level_Level, '-', 1)-2)
  3. Relevel: concatenate(T, ' - ', R)
  4. Difference: -Estimate
  5. Lower: -Upper_CI
  6. Upper: -Lower_CI
Cosmetics:
  • Filter: Exclude
    ⦿ Level_Level
    ⦿ Estimate
    ⦿ Lower_CI
    ⦿ Upper_CI
    ⦿ R
    ⦿ T
  • Properties
    Relevel => Level_level
    Difference => Estimate
    Lower => Lower_CI
    Upper => Upper_CI

[image]



PS: You selected ☑ E-mail notification if there has been a reply to this message
However, both replies bounced from the E-mail server (xxxxxxxx wasn’t found at yyyyyyyy.com. 550 5.1.10 RESOLVER.ADR.RecipientNotFound; Recipient not found by SMTP address lookup).
Since you registered in 2011 with your company’s e-mail address: Did you change your job? If yes, consider updating your profile (once logged in, click on msmnainar in the upper right hand corner [image] Edit User data E-mail click on [ Change e-mail address ]) or refrain from further E-mail notifications. THX.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
ElMaestro
★★★

Denmark,
2017-01-20 21:12
(2624 d 12:17 ago)

@ Helmut
Posting: # 16992
Views: 4,999
 

 Lexical order

Hi Hötzi,

Both software packages by default give the LSM difference in lexical order. If you coded test with T and reference with R, the difference will be given as R – T (since R < T).


Hey I did not know this :-)

Perhaps this explains why many companies are still using A and B for treatments:cool::ponder:

Pass or fail!
ElMaestro
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
103 visitors (0 registered, 103 guests [including 8 identified bots]).
Forum time: 09:30 CET (Europe/Vienna)

With four parameters I can fit an elephant,
and with five I can make him wiggle his trunk.    John von Neumann

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5