Bioequivalence and Bioavailability Forum

purvaph
☆

India,
2012-07-04 07:45
(5098 d 22:44 ago)

Posting: # 8884
Views: 9,302

To learn something new in 2×2 Randomization [General Statistics]

Hi All,

I have a query regarding 2*2 Crossover Bioequivalence

proc plan seed=0011202 ;

factors block= 2 ordered Subject=14 ordered;

treatments Sequence= 14 random;

output out=first

sequence nvals=(1 1 1 1 1 1 1 2 2 2 2 2 2 2)

random;

run;

Initially the output generated for randomization of 28 subjects starts from Subjects 1 to 28 with respective blocks and sequences.
Can I know the SAS code which will give me the randomization output from Subjects 21 to 28 without change in sequence and blocks?
Kindly let me know where those SAS Code should be added in PROC PLAN or in different step?

Regards,
Ajitha Ponnukuttan

d_labes
★★★

Berlin, Germany,
2012-07-04 19:06
(5098 d 11:23 ago)

@ purvaph
Posting: # 8891
Views: 8,065

Ugly Proc Plan

Post reply

Dear Ajitha!

Nothing simpler than that.
Try:

*the option /noprint prevents output which is annoying at least

output out=first names the output dataset;

Proc plan seed=0011202;

factors block= 2 ordered Subject=14 ordered /noprint;

treatments Sequence= 14 random;

output out=first

sequence nvals=(1 1 1 1 1 1 1 2 2 2 2 2 2 2) random;

run;

*number consecutively;

Data first;

  set first;

  subject=_n_;

run;

Title Randomization for subj. 21-28;

Proc print noobs;

  where subject>=21;

run;

This is an example of great creativity of the SAS developers.
Usually one has to use the noprint option on the Proc line (Proc plan seed=12345 noprint;).
But here this gives an error :angry:

—
Regards,

Detlew

purvaph ☆ India, 2012-07-24 09:11 (5078 d 21:18 ago) @ d_labes Posting: # 8973 Views: 7,821	Ugly Proc Plan Post reply
	Hey Thanks for the response...The Program actually worked..... Regards, Ajitha Ponnukuttan

jag009
★★★

NJ,
2012-09-18 19:50
(5022 d 10:39 ago)

@ d_labes
Posting: # 9237
Views: 7,544

Ugly Proc Plan

Post reply

Dear D_labes,

Do you see anything strange on my 4-way crossover randomization code? Just checking :-)

proc plan seed=12152;

      factors Subject=24 ordered Period=4 ordered;

      treatments Treatment=4 cyclic;

      output out=LatinSquare

             Treatment cvals=('A' 'B' 'C' 'D') random;

   run;

      proc sort data=LatinSquare out=LatinSquare;

      by subject period;

   proc transpose data= LatinSquare(rename=(period=_NAME_))

                  out =tLatinSquare(drop=_NAME_);

      by Subject;

      var Treatment;

   proc print data=tLatinSquare noobs;

   run;

Thanks
John

d_labes
★★★

Berlin, Germany,
2012-09-19 18:46
(5021 d 11:43 ago)

@ jag009
Posting: # 9239
Views: 7,629

Proc Plan 4x4 crossover

Post reply

Dear John,

❝ Do you see anything strange on my 4-way crossover randomization code?

Looking at the result of your code (hope you have checked it by yourself) I noticed that there is not too much random in your randomization :-D

.
Just a 6 times repetition of the sequences


1   ADBC

2   DBCA

3   BCAD

4   CADB

May be that some regulators/assessors are not satisfied with that. See Helmut's post on a similar question for a randomization of 2x2 crossover designs.

Try to use random instead of ordered after your factor Subject

—
Regards,

Detlew

jag009 ★★★ NJ, 2012-09-20 23:20 (5020 d 07:09 ago) @ d_labes Posting: # 9241 Views: 7,398	Proc Plan 4x4 crossover Post reply
	Thanks d_labes

jag009 ★★★ NJ, 2012-09-25 20:07 (5015 d 10:22 ago) @ d_labes Posting: # 9251 Views: 7,376	Proc Plan 4x4 crossover Post reply
	Thanks D_labes (Part 2) »...I noticed that there is not too much random in your randomization . ❝ Just a 6 times repetition of the sequences ❝ `❝ 1 ADBC` `❝ 2 DBCA` `❝ 3 BCAD` `❝ 4 CADB` I see, you were referring to the DB, CA, AD... Below is better... 1 ABDC 2 BCAD 3 CDBA 4 DACB Thanks John

d_labes
★★★

Berlin, Germany,
2012-09-26 13:59
(5014 d 16:30 ago)

@ jag009
Posting: # 9262
Views: 7,771

4x4 crossover - Latin square or Williams

Post reply

Dear John,

:confused:

What does the poet want to tell us? (An old question from school days.)

I didn't criticized the chosen Latin square. IMHO it is not necessary to choose a Williams square which is said to be balanced with respect to 'first order carry over' since our usual evaluation does not consider such an effect. Beside the fact that 'first order carry over' is highly criticized (see Stephen Senn's book "Cross-over Trials in Clinical Research").
The use of Williams design is nevertheless often used nowadays because of historical reasons.

What I tried to say was that your code results in a 'random' list which was a simple repetition of the four sequences, i.e subjects 1-4 were assigned to the four sequences, subjects 8-12 to the same four sequences and so on.
Try the code


proc plan seed=12152;

  factors Subject=24 random Period=4 ordered;

  treatments Treatment=4 cyclic;

  output out=LatinSquare

  Treatment cvals=('A' 'B' 'C' 'D') random;

run;

to see the difference.

—
Regards,

Detlew

To learn something new in 2×2 Randomization [General Sta­tis­tics]

Ugly Proc Plan

Ugly Proc Plan

Ugly Proc Plan

Proc Plan 4x4 crossover

Proc Plan 4x4 crossover

Proc Plan 4x4 crossover

4x4 crossover - Latin square or Williams

To learn something new in 2×2 Randomization [General Statistics]