Date:         Sun, 7 Feb 1999 08:42:23 +0100
Reply-To:     palm@algonet.se
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Martin Palm & Linda Simonsson <palm@ALGONET.SE>
Organization: Hemma
Subject:      Re: Q: paired vs unpaied t-test
Content-Type: multipart/mixed;

Hi!
 
If you have two groups, eg two groups of patients one receicving active
treatment and
the other one placebo, and you want to test the difference in treatment,
 
then you should use a
2-sample t-test (PROC TTEST).
 
If you have one group, eg  you will test if a group of patients has
improved over time,
then you should calculate the difference (diff=after-before) for each
patient and then test the difference
with a 1-sample t-test (PROC UNIVARIATE).
 
Linda
 
Sung-Il Cho wrote:
 
> Hello,
>
> This is only remotely about SAS, but
> I'm looking for a reference that compares the power
> of paired and unpaired t-test, especially describing
> the conditions under which one is more powerful than the other,
> and vice versa. (for 2-group data)
>
> Any input would be greatly appreciated!
> Sincerely,
> Sungil.
 
[text/x-vcard]

begin:vcard
n:Linda Simonsson;Martin Palm
x-mozilla-html:FALSE
adr:;;;;;;
version:2.1
email;internet:palm@algonet.se
x-mozilla-cpt:;-1
fn:Linda Simonsson, Martin Palm
end:vcard