Date:         Tue, 4 Mar 2003 16:47:14 -0600
Reply-To:     Deborah Wentworth <Deborah.Wentworth@ACORNCV.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Deborah Wentworth <Deborah.Wentworth@ACORNCV.COM>
Subject:      kappa statistics
Content-Type: text/plain; charset="iso-8859-1"

Greetings -

I have a question I'm hoping that this esteemed group can help me with.

I am using PROC FREQ with the AGREE option to get a weighted kappa for an ordinal measure that can range from 1 to 4.  This measure is assessed by two people, and I'm comparing their results and quantifying the agreement between the two.

For one of my subgroups, one person had results ranging from 1 through 4 while the other person did not assign any 1s.  When I ran this, SAS didn't produce the kappa statistic because my table was not square (# columns did not = # rows).

Is there a way that I can force SAS to produce a square table even though the column count will be zero for one of the columns?  I hope this question makes sense....

Thanks in advance - Deb