Date:         Thu, 17 Oct 1996 08:23:03 -1000
Reply-To:     Chun Huang <chun@MANA.MEDSURF.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Chun Huang <chun@MANA.MEDSURF.COM>
Organization: Mana Institute
Subject:      Re: PROC LOGISTIC question
Comments: To: "Ioannis C. Dimakos" <idimakos@MAILBOX.SYR.EDU>

Date:          Thu, 17 Oct 1996 11:29:15 -0400
Reply-to:      "Ioannis C. Dimakos" <idimakos@MAILBOX.SYR.EDU>
From:          "Ioannis C. Dimakos" <idimakos@MAILBOX.SYR.EDU>
Subject:       PROC LOGISTIC question
To:            Multiple recipients of list SAS-L <SAS-L@VM.MARIST.EDU>
 
Fellow SAS-Lers
 
        A colleague is using PROC LOGISTIC with a model statement:
model y=x1 x2 x3 x4,
where y is group membership (0/1) and x1-x4 regular SES variables.  The
question to which we could not find an answer is: "Can SAS report not only on
the effect of var. x2 on the outcome "y", but on the effect of various levels
of x2 (1, 2, 3, 4) on y?"
 
X2 represents "college level" and takes one of 4 possible values:
(1=freshman,...,4=senior).
 
Any ideas, suggestions would be appreciated.
 
        Thanks.
 
        John Dimakos
 
John:
 
First, you may consider using 3 dummy variables to represent 4
possible college levels. Second, if you treat X2 as an ordinal
variable, you can use UNITS statement to calculate odds ratios
corresponding to increase in different unit (1,2, or 3).
 
Good luck.
 
Chun Huang, Ph.D.
Hawaii Osteoporosis Center
401 Kamakee St.
Honolulu, HI 96814
USA
Email: chuang@mana.medsurf.com
Phone #: 808-592-2635