Date: Thu, 6 May 1999 13:38:22 -0400
Reply-To: "Hudson, Spencer" <shudson@VIROPHARMA.COM>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: "Hudson, Spencer" <shudson@VIROPHARMA.COM>
Subject: FW: What method should I use?
Content-Type: text/plain; charset="iso-8859-1"
Of course, strictly speaking, it is now too late to decide "...what
differences/changes are of interest to you..." because GBH has already
looked at the data.
Cheers,
Spencer Hudson
-----Original Message-----
From: John Whittington [mailto:medisci@POWERNET.COM]
<mailto:[mailto:medisci@POWERNET.COM]>
Sent: Thursday, May 06, 1999 10:53 AM
Subject: Re: What method should I use?
At 10:36 06/05/99 -0400, Huntress, Gary B. wrote (in
part):
>Using proc mixed, glm with repeated statement, and glm
with manova
>statement it is clear that there are significant
changes among all of
>the measurements, but none tell where the statistically
significant
>changes are. How do I pinpoint the significant changes
using SAS?
The general answer is that you need to decide what
differences/changes are of interest to you and examine their
'significance' - e.g. using CONTRAST and/or LSMEANS statements. Unless
it's unavoidable, you probably would not want to look at 'all possible
differences/changes' with a design as complicated as yours, since the
penalty in terms of 'multiple test correction' would then be high.
Regards,
John
----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296
730225
Mediscience Services Fax: +44 (0) 1296
738893
Twyford Manor, Twyford, E-mail:
medisci@powernet.com <mailto:medisci@powernet.com>
Buckingham MK18 4EL, UK
mediscience@compuserve.com <mailto:mediscience@compuserve.com>
----------------------------------------------------------------
| 
