Date: Mon, 15 Aug 2011 09:39:53 -0700
Reply-To: Steve Denham <stevedrd@yahoo.com>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Steve Denham <stevedrd@YAHOO.COM>
Subject: Re: Proc plan
In-Reply-To: <201108151551.p7FAl9Fw006452@willow.cc.uga.edu>
Content-Type: text/plain; charset=iso-8859-1
I don't see a need to include "block" as a fixed effect. The following should be adequate:
proc mixed;
class tx block;
model y=tx ;
random intercept/subject=block;
I prefer the subject= form of the random statement, so that I can flip easily to:
proc glimmix method=quad;
class tx block;
model y=tx;
random intercept/subject=block;
Good luck!
Steve Denham
Associate Director, Biostatistics
MPI Research, Inc.
From: William Shakespeare <shakespeare_1040@HOTMAIL.COM>
>To: SAS-L@LISTSERV.UGA.EDU
>Sent: Monday, August 15, 2011 11:51 AM
>Subject: Re: Proc plan
>
>On Sat, 13 Aug 2011 15:54:24 -0400, William Shakespeare
><shakespeare_1040@HOTMAIL.COM> wrote:
>
>>I'm reading this article on proc plan
>>(http://www.lexjansen.com/pharmasug/2007/ad/ad07.pdf). The authors
>>postulate a 3 treatment arm study with 300 subjects. This is their
>>explanation and the proc plan to make assignments in a permuted block
>>randomization:
>>
>>"Consider the previous example (section 2.1) with three treatments and
>>sample size 300. If the block size is 6 then there are 90
>>possible permutations in which each of three treatments can appear with
>>equal number of time (i.e. twice)."
>>
>>6 things taken 3 at a time is 120 permutations, but when you restrict it
>>to those with a maximum of 2 repetitions of any treatment it lowers the
>>count to 90. I can't for the life of me remember how this is computed.
>>Can anyone refresh my memory?
>>
>>Also, if one uses all 90 permutations then 540 subjects would have to be
>>randomized. Since the present study only uses 300 doesn't that leave the
>>possibility that the reprenstation of some pairs, for example, might be
>>unbalanced, e.g. the permutations utilized might have treatment sequence
>1-
>>2 more often than sequence 2-3. Is this not generally considered to be
>>undesirable in clinical trials?
>
>Another thing I'm trying to confirm is the proper proc mixed code to
>analyze this design (3 treatments with a permuted block randomization, all
>at a single location). I'm thinking:
>
>proc mixed;
>class tx block;
>model y=tx block;
>random block;
>
>but I'm not usre it's correct. And I still can't figure out how the
>authors of the paper figured 90 permutations and whether it matters if all
>are used in the trial.
>
>
>
| 
