Date:         Tue, 14 Nov 2006 16:27:47 -0600
Reply-To:     "Swank, Paul R" <Paul.R.Swank@UTH.TMC.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Swank, Paul R" <Paul.R.Swank@UTH.TMC.EDU>
Subject:      Re: question on Comparing two regression lines
Comments: To: jn mao <jn_mao@YAHOO.COM>
Content-Type: text/plain; charset="us-ascii"

The interaction is a test of equal slopes between groups and the fact
that it is not significant is good since the compariosn of the groups
across doses can proceed once the homogeneity of regression slopes seems
tenable. Now drop the interaction and run the model with just X and
group. This is akin to analysis of covariance. if the group effect is
significant, that means the groups do differ and the lack of interaction
means that the group difference does not depend on the dose. Dose may
still be related to the outcome. Of course, statistical tests are no
guarantee of difference, but an indication that there is evidence to
support the research hypothesis. And all this presupposes that your data
are reasonable for this analysis.


Paul R. Swank, Ph.D.
Professor, Developmental Pediatrics
Director of Research,


University of Texas Health Science Center at Houston

-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of jn
mao
Sent: Tuesday, November 14, 2006 3:29 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: question on Comparing two regression lines

Hello SASLs,

  I have a continuous Y measured at 3 doses of drug A given (X) between
two groups. My boss asked me to compare if overall Y is significant
difference with 3 doses between 2 groups.  I did 'proc glm' to compare
the regression line of Y with X (x=1, or 2 or 3 for 3 doses) between two
groups. The interaction p value is >0.05. I did conclude that overall Y
is not significantly different with increasing dose of drug A between
two groups. Is that conclusion right?

  As my boss asked that the interaction is not significant only means
that the two regression lines are parallel, but the difference between
two parallel lines could be significant. He said there must be some else
analysis that could answer the question of overall Y is different
between two groups. Could someone help me on this question? Thanks much.

  Jane


---------------------------------
Cheap Talk? Check out Yahoo! Messenger's low PC-to-Phone call rates.