Janell,
Whether you use a kappa statistic or ICC will depend on the nature of
your data. If it is categorical in nature, then kappa is indicated.
However, if the data are on an interval scale, then you can use an ICC.
If you decide to use a kappa statistic, I have syntax on Jason King's
website at Baylor College of Medicine that will conduct analysis for any
number of raters and categories. The syntax there will provide
solutions for Bennett's S, Fleiss' generalized kappa (often misreferred
to as Cohen's kappa), the real Cohen's kappa, Conger's kappa, and Gwet's
AC1. Andy Hayes at the Ohio State University has similar syntax for
Krippendorff's alpha on his website, which uses a bootstrap approach.
If you would prefer to have syntax for a single kappa-type statistic
rather than the entire batch, please contact me offline and I will send
it to you.
Baylor College of Medicine Website
http://www.ccitonline.org/jking/homepage/Andy Hayes at OSU
http://www.comm.ohio-state.edu/ahayes/Good luck,
Brian
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of
Janell Mensinger
Sent: Monday, November 26, 2007 4:53 PM
To:
[hidden email]
Subject: inter-rater reliability
Hello everyone,
I am trying to figure out how to calculate a Kappa statistic or perhaps
some
other inter-rater agreement statistic such as an intraclass correlation.
I
have 40 raters who rated 6 cases as passing or failing using a
new instrument. I would like to get a statistic that gives me a measure
of
interrater agreement among these 40 doctors. With SPSS I only know how
to
get a kappa on 2 raters using the crosstabs function. Any suggestions
here?
Many thanks in advance.
Janell Mensinger
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
