Date:         Mon, 9 Feb 1998 07:16:27 +1000
Reply-To:     Bob Green <bgreen@DYSON.BRISNET.ORG.AU>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         Bob Green <bgreen@DYSON.BRISNET.ORG.AU>
Subject:      Comparing Cohen's Kappas
Content-Type: text/plain; charset="us-ascii"

Jeff wrote,
 
>Thanks, Bob for your response.  To clarify my original message, I am
>comparing each of 2 independent groups to a third using kappas at the
>measures of agreement.  The question then becomes- is the kappa obtained
>in the first comparison statistically diffeent from the comparison
>obtained in the second?  Assume that both kappas are statistically
>significant.  That is, they significantly exceed chance agreement.
>This is akin to comparing two Pearson's r's using the Fisher's Z
>tranformation.  Thanks again.  Jeff
 
 
I'm not sure if the situation regarding comparing two Pearson's r's using
the Fisher's Z tranformation is the same as comparing two kappas. I say this
because kappa is a 'transformed' value, e.g it is akin to a 'correlation',
however, corrected for chance agreement often resulting in a often much
lower value than is obtained by correlation.  I'll let someone else comment
further on this.
 
My point is that even if you found a means to determine whether your three
groups had statistically different kappas, e.g group A had a higher level of
agreement than group B or C, it would still be important that their kappa
was above chance agreement, albeit one group displayed a lower level of
agreement.
 
I hope I am not complicating things,
 
Bob Green