Date:         Mon, 3 Aug 1998 15:27:50 -0400
Reply-To:     "Grigsby, Jared" <JGrigsby@CAROLINAS.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         "Grigsby, Jared" <JGrigsby@CAROLINAS.ORG>
Subject:      Kappa Statistic
Content-Type: text/plain

Hi there. I am trying to calculate the kappa statistic for inter-rater
agreement between two raters using PROC FREQ. My problem occurs when I
have an mXn table rather than an nXn table. This occurs for me when
there are three choices from which the raters have to choose: 1, 2, 3.
Rater A chooses 1 sometimes, 2 sometimes, and 3 sometimes. But Rater B
never chooses option 3. This results in a 2X3 table, and PROC FREQ will
not calculate the kappa statistic.
 
Is there a way to force PROC FREQ to print a row with frequencies of all
0 for option 3 for Rater B so that we have a 3X3 table from which to
calculate the kappa statistic? Or is there another way to handle this?
 
Thanks for your help.
_________________________________________________
 
Jared Grigsby
Biostatistician
Carolinas Medical Center
704-355-8787
704-355-1880 (fax)
jgrigsby@carolinas.org