Date:         Fri, 17 Dec 1999 18:56:54 GMT
Reply-To:     machellew@MY-DEJA.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         machellew@MY-DEJA.COM
Organization: Deja.com - Before you buy.
Subject:      Re: analysis of 4 period cross-over design (placebo controlled,
              dose respose)

Dear Mr. Joshy:

It would seem to me that finding a way to save your trial given the
difficulties you have encountered is more a problem in the domain of
epidemiologic methodology. I would therefore recommend that you post
this interesting question to EPIDEMIO-L: The Epidemiological Methods
List Server

It is a list server for the discussion of methodological issues
in the field of epidemiology. It also features current events, meetings,
and job postings.

TO SUBSCRIBE TO EPIDEMIO-L Send an e-mail letter to:

             LISTPROC@CC.MONTREAL.CA
with the command:

       ----> SUBSCRIBE EPIDEMIO-L your name
in the body of your e-mail letter


I look forward to seeing your post and the ensuing discussion on
Epidemio_L. Good luck!

Machelle Wilchesky
Division of Clinical Epidemiology
Mcgill University Health Centre
Royal Victoria Hospital Pavilion
Montreal Quebec
Canada
In article <9779D242E943D311AF1E0090278CABD4037951@ntsg46.novo.dk>,
  "GJOS (Grace Joshy)" <GJOS@NOVO.DK> wrote:
> Hello,
>
> I'm trying to analyse a 12 patient randomised placebo controlled
> dose-response trial. The trial had 4 treatments administered on four
test
> days to each patient. It was intended to be latin-square design with 4
> different sequences, each repeated 3 times. The primary endpoint is
log(auc)
> of 4 hr blood glucose.
>
> But something went wrong in the randomisation procedure and the there
are 12
> different sequences with no blocking in the trial now.
>
> ABDC
> ADBC
> ACDB
> CABD
> CBAD
> CDAB
> BDCA
> BCDA
> BCAD
> DABC
> DBAC
> DBCA
>
> Any ideas for analysis that could save the trial?
>
> Is something like the SAS code below valid?
>
> proc mixed data=trial order=internal;
> class subj_id dose ;
> model lauc = dose/e;
> repeated / type=cs sub=subj_id r ;
> contrast 'Placebo vs Active'
>    dose
>    -3   1   1   1   ;
> estimate 'Active vs. Placebo' dose -1 0.33 0.33 0.33/ cl;
> lsmeans dose/e cov cl;
> make 'lsmeans' out=ls_mod2;
> run;
>
> Thanks & regds,
> G.Joshy
>


Sent via Deja.com http://www.deja.com/
Before you buy.